Conference 7.286::digital

From:	RANGER::JCAMPBELL "Digital - the commodity PC software vendor  13-Jul-1992 1226" 13-JUL-1992 12:29:09.08
To:	@YAH.DIS,@YAE.DIS
CC:	
Subj:	Windows-NT Developers Conference - trip report and analysis

From:	ERIS::CALLAS "You have to respect someone who can spell TUESDAY, even if they can't spell it right.  13-Jul-1992 1121" 13-JUL-1992 11:21:03.82
To:	BRONS,RANGER::JCAMPBELL
CC:	
Subj:	Barry at the NT conference

From:	3D::NISHIMOTO "Pete Nishimoto MLO1-2/U2 223-6343  13-Jul-1992 1043" 13 July '92 10:39 am
To:	ALPHAS PEXERS 
CC:	
Subj:	 FYI - B. Tannenbaum's trip report form the NT developers conference

From:	SITBUL::FLATTEN "13-Jul-1992 1009" 13-JUL-1992 10:17:08.45
To:	@DIS:EXT_STAFF
CC:	STEVE,SAUL
Subj:	FYI - B. Tannenbaum's trip report form the NT developers conference

From:	WRKSYS::DSSDEV::DSSDEV::TANNENBAUM "Any sufficiently advanced magic is indistiguishable from technology  13-Jul-1992 0936 -0400" 13-JUL-1992 09:38:09.39
To:	@MICROSOFT
CC:	
Subj:	Win32 Developers Conference Trip Report

Trip report for Microsoft Win32 Developer's Conference

Summary:

Windows-NT is a *real* operating system.  It features a full 32-bit virtual
address space, pre-emptive scheduling and multiple tasks.  The feel of the
presentations is that this is a re-implementation of VMS in C and C++.  Anybody
who saw the Mica specs will recognize the object-oriented flavor of this OS.

While it was amusing to watch the MS-DOS developers struggle to get their minds
around concepts that we take for granted, such as multiple virtual address
spaces and pre-emptive scheduling, they will master these ideas *very* quickly.

Windows-NT provides a number of opportunities for Digital, but we will have to
move fast to capitalize on them.  In a lot of ways, we've solved many problems
that the PC world is just starting to recognize.  However, Digital has
little-to-no name recognition in the PC space.  Ask a PC user about networking
and he'll talk about Novell.  If we want to compete in this area, we're going
to have to learn how to sell our products.

3D graphics can be a major opportunity.  Microsoft is working with SGI to
"integrate GL into the GDI (graphics library)" but claimed that GL will not be
available until V2.0.  Other points to consider:

   � SGI was showing GL on a Crimson, and it was not blindingly fast, even
     with some acceleration.

   � Ithica Software was at the vendor's area, but just handing out
     literature.  However, They were asking questions at the sessions
     indicating that they either already have or are developing a Windows
     implementation of HOOPS.

We have an opportunity to establish PHIGS in the Windows-NT market, but we're
going to have to act fast.  I believe that the window will start to close when
V1 of Windows-NT is released in December.

In order to do this, we will have to make the following changes to PHIGS:

   � Full shading, lighting and Z-Buffering in software since there's no
     hardware support.  This should have been done long ago anyway.  We
     should be able to grab most of this from software PEX.

   � Windows graphics device handler.  We *may* be able to get by with a pixmap
     rendering implementation (no input) for first release.

   � Re-organization of PHIGS so we don't carry around baggage we'll never use
     like painter's sort.  This needs to be done anyway.

   � True central structure store. This needs to be done anyway, but I believe
     that it's not required for first release on Windows.

The long version:

Whenever I go to California, there's always something that remind me that I'm
in a "different" part of the country.  This trip, they all occurred on the
weekend.

I spent the 4th of July with relatives who live in Larkspur, about 20 minutes
off the Golden Gate bridge in Marin.  We went to see the traditional 4th of
July parade, complete with politicians, marching bands, cub scouts... and
belly dancers.

We ended up at my cousin's house for a block party.  One of her neighbors had
been to a Club Med at Ixtapa, Mexico, where they had a flying trapeze.  He
liked the flying trapeze so much that he ripped out his backyard and
constructed one.  He'd invited all of the neighbors in to try for the party. 
Unfortunately, dinner was called before I got to try.

Sunday, I went and played tourist over at Fisherman's Wharf, and then headed to
an art and Jazz festival at Filmore street.  I was about two-thirds of the way
around the festival when I came across a crowd in front of two performers who
called themselves "THOTH" (with the "o" as a yin/yang sign).  The performer in
the back was muscular, tanned, not wearing much more than a loincloth, and 
posed in front of a large drum.  The second performer was wearing a loincloth,
a vest and a pair of of black fishnet stockings.  He was clutching a violin to
his chest and posed like something from a "Conan the Barbarian" novel cover
using the violin bow as a sword.  Then the drum player started doing Tai Chi
exercises with his drum.  I said to the woman standing next to me that I wished
that I'd brought my camera, since no one back home would believe this...

Rather than give you a session-by-session summary, I'll simply go over what I
considered the highlights.  If you want details on the sessions, I have the
complete set of session notes in my office.

      � Digital came off *very* well throughout the conference.
	   - It seemed like half of the machines in the demo room were
	     DEC 450STs.
	   - Digital was mentioned numerous times during the conference.
	   - The Alpha-NT presentation was well attended and well received.
	   - The Digital/Microsoft party at the Exporatorium was extremely
	     well handled.

      � I asked the folks at the Borland booth (*the* pre-eminent C/C++ vendor
	in the PC space) about their plans for "non-Intel architectures". 
	Their response was that they found the MIPS architecture boring, but
	were very interested in Alpha and wanted to talk with us about the
	GEM back-end.  Ben Schreiber is apparently putting them in contact with
	the folks in TLE.

      � Many of the people coming from a DOS world were a bit overwhelmed
	by the concepts being presented - multiple virtual address spaces,
	pre-emptive scheduling, asynchronous completion routines, etc.
	However, I expcet that they'll understand this brave new world quickly.

      � Microsoft is not documenting any of the executive interfaces (which
	had a lot of neat stuff).  For V1, all applications must use one of
	the following subsystems:
	   - Win32 (fully integrated with the graphics)
	   - POSIX (character cell only)
	   - OS-2 (character cell only)
	You cannot mix the subsystems.

      � Microsoft clearly does not consider the workstations vendors (either
	Unix or VMS) to be much of a threat.  Bill Gates mentioned in the Q&A
	session that followed the his keynote address that Microsoft is one of
	the largest Unix vendors (XENIX) and that the Unix market is 1% of the
	PC market.  Many of the Unix developers requested Xlib support and
	were told that Microsoft had no plans to supply one.

      � One of the recurring phrases from the presenters was "... when we were
	at Digital ..."

      � Microsoft has put itself into a win-win position.  If Unix takes off,
	they have XENIX.  If OS-2 takes off, "we wrote 95% of OS-2.  We don't
	need help from IBM."

      � The GDI (graphics library) has been moved into a separate process
	from the application for robustness.  Like Xlib, they've implemented
	batching (buffering).  This has a major performance impact on moving
	large pixmaps.

      � Microsoft C has some *very* nice exception handling features. 
        Basically, exceptions can be handled on a block-by-block basis.  We
	should look into getting this into our C implementations.

      � Microsoft employs one tester for every two developers!!!  One group
	was described as 120 developers, 70 test writers and 20 test runners.

      � Microsoft/Test is based on "Test Basic".  All of the input stuff that
	we've been worried about is handled, since you've got a full language
	at your disposal.

Microsoft has learned from it's experience with OS-2.  In order to convince
software developers that they should move to Window-NT, they have redefined the
existing Windows API as "Win16", and an expanded Windows API known as "Win32". 
Both APIs are supported by Windows V3.1 and Windows-NT.  Migration from Win16
to Win32 looks pretty straightforward (of course, I don't have any existing
Windows code to worry about)

Windows is evolving into a family of products:

	Windows V3.1 - The DOS version, currently shipping

	Windows-NT V3.1 - Cutler's new operating system with Windows V3.1 as
		it's GUI.  Expected to FRS in December.

	Windows for Workgroups V3.1 - DOS based.  Includes networking tools.
		Expected to FTS in "early '93".

	LAN Manager for Windows NT V3.1 - NT based.  Include network control
		tools.  Expected to FRS "late '92"

Beyond the current products is a project called "Cairo" which is an object
oriented environment.  The goals sound good, but they lost me when they started
describing this in terms of OLE (Object Linking and Embedding) since I don't
know anything about OLE.

From:	RANGER::SPRUNG "It's always something  06-Jul-1992 2101"  6-JUL-1992 21:03:03.60
To:	@\dis\vns
CC:	
Subj:	MS-Windows app development environment


----- Begin message from MSEDEV::MUNROE  6-Jul-92
    
From:	MSEDEV::MUNROE "Rebecca Munroe 232-2374  06-Jul-1992 1639"     6-JUL-1992 17:00
To:	@GEN
CC:	MUNROE
Subj:	microsoft eassse-of-use design factors

From:	ISLNDS::MALTZMAN "06-Jul-1992 0812"  6-JUL-1992 08:33:18.40
To:	@CA.
CC:	MALTZMAN
Subj:	CUSTOMER SAT COMPETITIVE FYI:  MICROSOFT EASE-OF-USE DESIGN FACTORS

    Microsoft introduced an application design guide for windows user 
    interfaces.    Contains artistic guidelines, library of icons and 
    cursors that can go right in applications,  and terminology in 13 
    languages.  Looks like this might become a big item..........Alan


REDMOND, Wash.--(03-JUL-92 BUSINESS WIRE)--Microsoft Wednesday announced a 
family of products that will help foster consistency and ease-of-use 
among applications designed to run in the Microsoft Windows 
environment. 

The new user interface design tools include: 

-- A 220-page application design guide that provides instructions
for designing user interfaces. 

-- A set of companion disks that offer an example application, 
artistic guidelines for designing graphical elements of the user 
interface, and a library of icons and cursors that can be 
incorporated directly into applications. 

-- A Microsoft University course on user interface design. 

-- A terminology reference that provides user interface terms in 
13 languages 

``We've gotten a great many requests from developers for help in 
designing user interfaces for Windows-based applications,'' said 
Jonathan Lazarus, vice president of systems strategy at Microsoft. 
``This family of products provides a complete set of tools and 
guidelines for designing user interfaces.  This lets developers build
applications that fully exploit what we've termed `The Windows 
Interface.''' 


WHAT IS `THE WINDOWS INTERFACE?'
--------------------------------

The Windows Interface is what the user interacts with and 
experiences while using the Windows environment.  That interaction is
dependent on the underlying Windows technology, but more importantly,
how software integrates those technology factors into the user 
interface design. 

There are five major design factors that user interface designers
can choose to exploit: data integration, input/output interfaces, 
consistency, ease-of-use, and scalability. 

Ideally, future applications for Windows will incorporate all 
these design factors and take advantage of the supporting 
technologies that the Windows environment provides. 

The Windows Interface tools and information will help developers 
address consistency and ease-of-use design factors, which are all 
imminently important issues in computing today. 

Though Microsoft is committed to helping developers write more 
consistent applications for Windows, Lazarus pointed out that the 
consistency guidelines are totally voluntary.  In fact, there may be 
some recommendations that do not apply to a certain applications. 
Our goal is to provide a suite of user interface products to use as a
baseline when designing a Windows-based application. 


THE WINDOWS INTERFACE:  APPLICATION DESIGN GUIDE
------------------------------------------------

A key element of the user interface tool set is The Windows 
Interface: An Application Design Guide.  This 220-page book provides 
extensive guidelines for designing user interfaces for Windows-based 
applications.  It includes both general principles and methodologies,
and specific recommendations for menu items, screen elements, dialog 
boxes and more. 

It is available today in bookstores at a suggested retail price 
of $24.95, and may be ordered directly from Microsoft Press by 
calling 800-MSPRESS.  In Canada, call Macmillan Canada at 
416/293-8141.  It is also shipping with the Microsoft Windows 
operating system Version 3.1 Software Development Kit. 


COMPANION DISKS
---------------

A set of companion disks to the ``Windows Interface: An 
Application Design Guide'' is also available.  These disks include 
the following materials: The Windows Interface: Interactive Design 
Guide The Windows Interface: Interactive Design Guide, developed by 
Microsoft Consulting Services, is a demonstration application that 
incorporates the ideas and recommendations presented in ``The Windows
Interface: An Application Design Guide.'' It also includes an on- 
line version of the manual's contents with powerful, full-text 
searching capabilities for instant access to all descriptions the 
user is interested in locating. 

When the sample application is run in information mode, clicking 
on any user interface item will bring up the explanation of that item
from the Application Design Guide.  When the application is run in 
demonstration mode, clicking on an item will cause it to behave just 
as it would in a normal application. 

It's also possible to launch the on-line reference manual 
directly. 


THE WINDOWS INTERFACE:  VISUAL DESIGN GUIDE
-------------------------------------------

The Visual Design Guide is an on-line reference that complements 
the information in ``The Windows Interface: An Application Design 
Guide.' It focuses on artistic considerations, such as how to use 
color effectively, how to design icons, and how to make buttons look 
raised, depressed, and inactive.  It also contains examples of 
visuals, including toolbars, icons and different types of buttons. 


THE WINDOWS INTERFACE:  BUTTONS AND CURSORS
-------------------------------------------

The Windows Interface: Buttons and Cursors is a library of images
that developers can incorporate directly into their applications.  It
includes button images for such commands as Cut, Copy, Paste, Open 
File, New File, File Close, Save, Print, and Help, and additional 
cursors such as the hourglass, arrow, cross-hairs, and I-beam.  The 
disk includes license agreements, allowing developers to use the 
images in their products royalty-free. 

The disk also comes with a dynamic link library (DLL) that can 
create all states of a button (ex.  raised, depressed, inactive) from
the surface area of the button.  By shipping their applications with 
this DLL instead of bitmaps for each button in each state, developers
can significantly reduce the size of their applications. 

The DLL will also work with any custom button images the 
developer creates. 

An example application included on the disk demonstrates how the 
DLL is used in applications.  Source code is provided for both the 
example application and the DLL, allowing developers to see the logic
that is used and modify it if desired.  The license allows the DLL to
be distributed in modified form.  The disk also includes an on-line 
document that explains how to use both the DLL and the example 
application. 

Users may download Buttons and Cursors from CompuServe on the 
Microsoft forum.  Microsoft is making this product available at no 
charge to encourage consistency in the use of buttons and cursors. 

The above user interface design tools may be licensed by software
vendors as part of the Open Tools License program.  The companion 
disks are available today at a suggested retail price of $24.95 and 
can be ordered directly from Microsoft Press. 

Registered users who purchased the book will be offered a special
upgrade price of $14.95 for the companion disks. 

A future edition of the Application Design Guide will come with 
the disks and will have a suggested retail price of $39.95. 


MICROSOFT UNIVERSITY COURSE
---------------------------

Microsoft University recently introduced User Interface Design 
for Microsoft Windows 3.1, a one-day course, to their programming 
curriculum.  This course complements the material in ``The 
Application Design Guide,'' providing more detail on the principles 
and methodologies involved in creating effective user interfaces, 
along with information on testing applications for ease of use. 
Courses from Microsoft University are available through MSU Training 
Alliance members, customer onsites and through a variety of Microsoft
University locations; interested parties should call Microsoft 
University at 206/828-1507 for details. 


TERMINOLOGY REFERENCE
---------------------

The final tool in the user interface product set is a terminology
reference guide that is still in production.  This localization guide
contains definitions, plus equivalent terms, in 13 languages for such
user interface terms as scroll bar, title bar, minimize and maximize.
More than 400 user interface terms are included.  The terminology 
reference will be provided in both print and on-line form.  It will 
be available from Microsoft Press this fall. 


SYSTEM REQUIREMENTS
-------------------

All of the above design tools require the use of the Windows 
operating system version 3.1, which requires MS-DOS Version 3.1 or 
later (version 5.0 or later recommended); a PC with at least a 80286 
microprocessor (80386SX or higher recommended); and 640K conventional
plus 256K extended memory.  A pointing device is strongly recommended
along with a 5.25-inch high-density or 3.5-inch disk drive and a hard
disk with 6MB available (10MB recommended); EGA, VGA, Super VGA, XGA,
8514/A, Hercules* graphics card or compatible video graphics adapter 
and monitor support by the Windows operating system (color VGA or 
better recommended). 

Founded in 1975, Microsoft (NASDAQ:MSFT) is the worldwide leader 
in software for personal computers.  The company offers a wide range 
of products and services for business and personal use, each designed
with the mission of making it easier and more enjoyable for people to
take advantage of the full power of personal computing every day. 


NOTES: 

Microsoft and MS-DOS are registered trademarks and Windows is a 
trademark of Microsoft Corp. 

CompuServe is a registered trademark of CompuServe Inc.

Hercules is a registered trademark of Hercules Computer Technology

All prices listed are U.S. suggested retail prices.


----- End forwarded message

I think that Windows NT will quickly take out the workstation vendors as we
know them, and bridge desktop PCs to high powered servers with a common OS.

However, I think the Apple Newton is probably more representative of "the next
wave". Windows NT doesn't get you to consumer electronics comfortably, and
that's where the "Interpersonal Digital Connection" market will be. Just
imagine a worldwide implementation of Notes for Newton, Mail for Newton,
TeamRoute for Newton etc without being limited to any one enterprise's network
infrastructure... now that'd be a pretty large market. And i'd buy one.

								- Ian W.

Re: .0   Small Nit - "Catch a Wave"  Jan & Dean

    Re: .4 Thanks for historical correction.
    
    							Jon

From:	TLE::AMARTIN "Alan H. Martin  14-Jul-1992 1257" 14-JUL-1992 12:56:06.86
To:	JLC
CC:	
Subj:	1 of 2

From:	SR::MURPHY "Sara Murphy, DTN: 381-2052  12-Jul-1992 2234"   12-JUL-1992 22:36:36.37
To:	@C_TEAM
CC:	
Subj:	NT Developers' Conf Q&A - C/C++ questions are at the end

From:	TLE::CONTI        "Bob Conti 603 881 2086 Just DO It"  8-JUL-1992 07:17:10.74
To:	SMURPHY,FOSTER,GOLLY::KNIGHT
CC:	
Subj:	NT Pro. Dev. Conf., Day One Q&A

From:	TLE::CONTI "Bob Conti 603 881 2086 Just DO It  08-Jul-1992 0713"  8-JUL-1992 07:16:50.02
To:	@[CONTI.DIS]PCDEVO.DIS
CC:	
Subj:	NT Pro. Dev. Conf., Day One Q&A

From:	DECWRL::"[email protected]" "Richard A. Wells"  7-JUL-1992 11:05:28.49
To:	me-dec <[email protected]>, nancy-b <tle::beckley>, bob <tle::conti>, Dieter <4gl::heinzer>, kathleen <aimhi::bailey>, tom <tuner::bailey>
CC:	
Subj:	NT Pro. Dev. Conf., Day One Q&A

Question and Answers in Main and Selected Breakout Sessions, Day One
Microsoft Windows Professional Developer's Conference
6 July 1992, Moscone Center, San Francisco

Reported by Richard A. Wells (TLE::WELLS)

[For all of these talks the slides are in the proceedings which I
will be bringing back with me.  I have only paratranscribed the
question and answer sessions from each of these talks.]

In this issue:
  Win32 Base APIs
  Win32 GDI (Graphics Device Interface)
  Window Manager API (USER)
  SQL Server
  C++ Exception Handling


Win32 Base APIs
---------------

Q: [Dave Solomon, formerly of VMS development] I don't see any
generalized lock management services?

A: You're right, we don't have that level in the first version.


Q: Can you have multiple views on a single file map?

A: Yes, you can have multiple ones of different sizes, with different
read/write access.


Q: You mentioned both structured exception handling and status return
codes.  Can you use just one?  Or did you consider building an API
that only uses the exception mechanism?

A: We thought about that, but it would be a nightmare.


Q: Can the scheduler be replaced by third-parties for real-time
applications?

A: We aren't doing that yet, but if you are really interested, please
talk to us or submit requirements to product management for version
2. 


Q: What does it mean to use threads and run on 3.1?

A: The Win31s libraries will not crash on 3.1, but the calls that
don't make sense will return error codes.  So your app can be aware
of the platform it is running on.


Q: Some people have trouble deciding how much shared memory to
allocate.  Can that be done dynamically?

A: Strictly speaking, no, but that doesn't stop you from creating
multiple shared memory sections dynamically.


Win32 GDI (Graphics Device Interface)
-------------------------------------

Q: Do you actually copy all of the data between client and server
components? 

A: Yes, into the shared buffer.


Q: Do you see any perf issues (benefits lost) with the current batch
of accelerator cards? 

A: We've tried to pick a set of calls/device driver requests that the
cards can handle well.  I think it will work well with these devices.


Q: Have you tested client-server caching over a phone line.

A: Remote calls to GDI is not supported in this version.


Q: Does it work with 24-bit color?

A: Yes.


Q: Is there a demo here of this?

A: I don't know.


Q: Will the new GDI support multi-head, multiple-display cards?

A: Yes, but not in the same way as the Macintosh.  Only one device
can have the desktop.  So, if there is a device-driver that
understands all of your cards, it can manage this.


Q: Would you implement an X server as a client or a device driver?

A: Client.  Performance should be quite acceptable.


Q: When [SGI's] OpenGL is integrated, where?

A: Not as a separate subsystem, but as part of the GDI.

[SGI is demonstrating OpenGL running under NT.  One demo showed a
window with a 3D representation of an undulating planar surface.  It
was very quick and smooth.  If NT builds this in as a standard
component, NT will probably make even more inroads into Unix
workstations' (previous) domain.]


Q: When we implemented our drawing with the PolyLine call we
discovered certain degenerate cases.  Will those types of cases also
exist in Paths, and will you document those?

A: You should talk to me directly with specifics.


Q: Will the Win3.1 device driver that I have for my graphics
accelerator card be obsolete for NT?

A: Yes.  You will need completely new drivers.

Q: Are manufacturers working on those?

A: I think so, but I don't know what amount of coverage we have.


Q: [Some question about memory use.]

A: One shared memory buffer per THREAD that communicates with the GDI
server.  But it is only about 8K/thread.


Q: In constructing metafiles what are the implications of having
complicated transforms?  Will all details be taken care of?

A: Yes, it should be.

Q: Can I do that as I CREATE the metafile.  [Complicated question
about paths and transforms.]

A: You'll have to read the path API documentation.


Window Manager API (USER)
-------------------------

Q: Is there any future plan to make Win32s pre-emptive?

A: Yes, in fact.  I'm not the qualified person to answer that, but I
think the next versions of DOS and Windows will do this.


Q: Does Win32s have Unicode support in it?

A: I don't know.  I won't answer that.


Q: Do you plan to implement a StyleTextEdit control as the Mac has
it? 

A: We've had a lot of requests for that.  We make new features
decisions based on ISV feedback.  That one has been a high one, but I
can't say for sure it will be in the next release because we haven't
tabulated it.


Q: If an application explicitly loads a DLL and the EXE crashes the
DLL will be left in memory and perhaps in an indeterminate state, in
Window 3.1.  Will this happen in NT?

A: No, the models are completely different.  If a process goes away
so will its DDLs.


Q: [Something about pre-emptive input model on Win3.1]

A: Win32s on Win3.1 will use the same model as is used today.


Q: What should I do about the change from SetFocus to
SetForegroundWindow?  Reason is if you have an alternate input device
and you need to configure it to be specific to the application that
is currently using.  Presumably you can use Set/GetForegroundWindow
to find out about the application to modify the device?

A: One thing to know about calling operations on other windows is
that if you call USER there will be synchronization with the other
process, because of the callback model.  If you do this to a window
whose thread is hung you may wait for a long time...


Q: One of the big features two years was backing store.  What is the
state of that?

A: A future release feature.  We decided it wouldn't get done in
time. 


Q: What about support for tablets, switching between relative and
absolute mode.  The situation is really bad today.  We'd like
Microsoft to define a standard interface.

A: I've heard about that a lot.  Talk to your development consultant.

Q: We've done that.

A: OK, I'll have to look into it.


Q: In 16-bit Windows text controls limited to 64K.  Has this gone
away? 

A: Yes.

Q: In preliminary versions, string copy functions are limited 64K.
Is this still true?

A: Printing holdover from 16-bit version.  This is NOT still true.


Q: What about support for multiple keyboards and mice?

A: Not in this first release.  If it is really important, get in
touch with your ISV relations contact or talk to me directly.


Q: Could you expound a bit on the message queues of modal vs.
modeless dialog boxes?

A: Yes.  For one thing, we don't support system modal windows.


Q: I develop applications for the public.  Is it possible to develop
an application that users can NOT get out of.

A: Yes, but there are many things to consider.  Not as easy as it was
using a system modal window.  However, you can trap all events.


Q: I would assume that a modal dialog box only gets messages on its
own thread.  Is that true?

A: Yes.

Q: So it would only block that thread?

A: Right.


Q: Do the various string macros automatically cast themselves
correctly for using Unicode?

A: Yes.

Q: What about string resources?

A: You just need to call the right API.  The resources are STORED as
Unicode, you retrieve them in the form you want.


Q: If an application is using Windows DDE messages directly, how is
backwards compatibility going to work with shared memory?

A: We've implemented a robust DDE mechanism that records ALL
communication and ensures that all the right events get put in the
right application's buffer.


Q: If A is a good app, B is a bad app.  B goes into disk I/O.  User
clicks around B.  Nothing happens, user switches to A.  Finally B
wakes up.  Does it pop to the front?

A: No. It receives the message but it doesn't become the foreground
window. 

Q: So it doesn't do a SetFocus?

A: Yes, it will, but it won't become the FOREGROUND window.


SQL Server
----------

Q: You mentioned that the ODS layer is fully compliant with SQL
Server.  I thought there were some hitches?  Specifically "compute by".

A: By making them integrated we are forcing ourselves to come to
grips with this.  We have to add them and we are adding them in one
place:  ODS.

Q: When you copy SQLSERVER.EXE from OS/2 you don't need the
associated DLLs?

A: Yes, you need the ODS DLL and there is a server-side NETLIB that
needs to be copied.  This does not include the client-side, which you
will need, too.  (More DLLs.)


Q: How about my current Win3.1 apps that talk to OS/2 1.x SQL
Servers? 

A: Win 3.1 clients will run UNCHANGED against an NT SQL Server.  In
order to run a Windows NT 32-bit client you will need an updated
[network] DLL.  You can even run the OS/2 SQL Server in NT in an OS/2
subsystem.  In any case you will need to load the appropriate DLLs
[on the server, I think].


Q: Have you run any TPS benchmarks?

A: We have run some, but not full TP1 benchmark.  It is definitely
faster than OS/2 and uses less CPU.


Q: How does your implementation differ from Sybase on, for example,
Mach. 

A: We are using NT threads to give us MP capabilities.  For example,
the Sybase implementation which requires extra work to take advantage
of additional processors.  We have single implementation which works
fine on a uniprocessor but get multiprocessing advantages through NT
threads automatically.


Q: What do I get under Win32s on a DOS platform?

A: I don't know.  I think there will have to be a different version
of the DLL under Win32s.


Q: Which is faster: Sybase on a Unix 486 or SQL server on NT?

A: I suspect we'll be faster because of architectural differences,
but I'm not sure.


Q: [pricing]

A: Pricing will be microcomputer, not minicomputer priced.


Q: Is SQL Server "throttlable"?

A: We intend to supply parameters that control how many threads are
allocated. 


Q: Is tempdb dynamic?

A: Yes.

Q: So it will contend with space for my procedures?

A: No, it doesn't come out of the space specified by your memory
parameter. 


Q: You seem to indicate TCP/IP will only be available through a
bridge.  Does this mean we won't get a TCP/IP NETLIB for NT?

A: That is not strictly true.  There are possibly server-side
NETLIBs. 

Q: I'm specifically interested in connecting NT clients to Unix
Sybase servers.

A: I see.  That will be in the Production version.

Q: Do we automatically get these systems?

A: Purchase through Internal Sales for $295.


Q: Is Sybase developing DB-LIBs for NT servers?

A: I assume you mean for TCP/IP?  Gary Voff is the product manager
for SQL Server.

A (GV):  Currently we share the DB-LIB development with Sybase.  We
do the development of the API for MS platforms, Sybase doing the
compatible implementations for non-MS platforms.


Q: When you mentioned released server on OS/2, which version do you
mean? 

A: SQL Server 4.2 can be run on OS/2 2.0 right now, but it is not
certified because we only just received copies of the network
software.  We are testing and it should be certified in a few weeks.


Q: Do the SA enhancements allow the diagnosis of a deadlock,
specifically which resource is being contended for?

A: That's a good idea and the extensibility is there to do that.

Q: Well, you have my vote.


C++ Exception Handling
----------------------

Q: What is in  C7.0?

A: Nothing.

Q: When will we see anything?

A: Not at liberty to say specifics, but something next year.

Q: What about under 3.1 or Win32s?

A: No.

Q: So there is nothing we can do?

A: There is support for C-style value exceptions in the NT PDK with
the 32-bit compiler.

Q: Is it your intention to be COMPLETELY compatible to ANSI spec?

A: Yes, but we will extend it compatibly to support resumably
semantics. 

Q: How will you merge terminate vs. resume semantics?

A: Under NT, in the C style, resumption is possible when the filter
is being executed.  Once you are in the accept clause objects have
already been destroyed.  I made a proposal to the ANSI committee in
1990 that is fully compatible.  Termination exceptions ALWAYS
terminate.  Resumable exceptions may terminate OR resume.


Q: Can you explain exception handling in MFC?

A: They provide library code that provides some of the flavor of C++
exceptions.  Once they have C++ compiler that truly supports
exceptions, they will be able to recompile their macros to take
advantage of this.


Q: C++ handling is very similar to what Ada does...

A: ... except type semantics, which is a BIG difference ...

Q: You suggested that something like "out of memory" should be an
error, not an exception, but the Ada and C++ exception handling
designers think it SHOULD be an exception.

A: I came from the Unix world and I know that it is a Unix
"arrogance" is that out-of-memory is something that happens rarely.
Under Windows and DOS with 64K segments that this happens all of the
time.  So it is really a case of programmer perception.

Q: Is Microsoft planning on supporting templates?

A: Absolutely!

Q: When?

A: Can't tell you, but it WILL happen for competetive reasons.


Q: Is the C++ base class problem being looked at and does Microsoft
have any interest in this?

A: Yes, I submitted it to the ANSI committee years ago.  Probably
won't see ANSI C++ until 2010, but you'll see an implementation in
Microsoft sooner.

Q: Can we get copies of your proposals.

A: I'll have to ask our legal team.


Q: Will throw and catch be available to other languages?

A: Yes, we disclosed the object model of our C++ code.  Our intention
is to do the same for our excpetion model.


Q: [Bob Peterson, DEC SDT]  How will this exception model affect
other languages apart from C++?

A: In the NT model, there is a language-handler attribute per frame
(and user can provide one per function).  By default this is a C
language handler.  The C++ compiler will generate C++ language
handlers.  NT itself doesn't care.  It just walks the frames and asks
the language handler about an exception through a particular
interface.  Today's language handler in the PDK is only for C.


Q: DLLs raising typed exceptions causes problems but you didn't
mention solutions.

A: I'm not at liberty to do so yet.  Most common situation is when an
application has normal classes and exception classes.  Virtual
functions provide a very complicated callback scenario.  The biggest
issue is when the client code throws an exception the DLL doesn't
know anything about it (unless it has a catch(...), which catches all
types).  The DLL can't mess this up.  Caller might have a catcher
higher up the call tree.  The DLL and the EXE have index sets for the
exceptions they are handling.  The CFRONT implementation proposal
uses mangled names of the types and search them until they find the
one they need.  And THIS is a little bit too expensive.  We have
defined an index scheme that is quite optimized, but I can't give you
the details now.



Distribution:
  me-dec >internet:[email protected]
  nancy-b >internet:[email protected]
  bob >internet:[email protected]
  Dieter >internet:[email protected]
  kathleen >internet:[email protected]
  tom >internet:[email protected]


% ====== Internet headers and postmarks (see DECWRL::GATEWAY.DOC) ======
% Received: by enet-gw.pa.dec.com; id AA17926; Tue, 7 Jul 92 08:02:45 -0700
% Received: by ihb.compuserve.com (5.65/5.910516)id AA02096; Tue, 7 Jul 92 11:02:38 -0400
% Date: 07 Jul 92 10:42:59 EDT
% From: "Richard A. Wells" <[email protected]>
% To: me-dec <[email protected]>, nancy-b <tle::beckley>, bob <tle::conti>, Dieter <4gl::heinzer>, kathleen <aimhi::bailey>, tom <tuner::baile
% Subject: NT Pro. Dev. Conf., Day One Q&A
% Message-Id: <[email protected]>

From:	TLE::AMARTIN "Alan H. Martin  14-Jul-1992 1257" 14-JUL-1992 12:56:27.03
To:	JLC
CC:	
Subj:	2 of 2

From:	SR::MURPHY "Sara Murphy, DTN: 381-2052  12-Jul-1992 2247"   12-JUL-1992 22:50:11.86
To:	@C_TEAM
CC:	
Subj:	NT Conference - Second day (includes the Alpha NT presentation)

From:	TLE::SURELY::will "Becky Will"  8-JUL-1992 12:20:32.61
To:	tle::beckley, peznqt::welch, tle::purcell, tools::rice, clt::trogers,
	tle::youngs, tle::benoit, kazak::clinkenbeard, tle::ellenberger,
	tle::foster, tle::karam, banana::klein, gitdwn::denise, tle::smurphy,
	will
CC:	
Subj:	comments from Supnik/Benn at NT conf on Alpha/NT langs

I've extracted appropriate comments 1st, then attached the whole trip
segment from Richard.
...
Q: Will DEC FORTRAN be ported to Alpha/NT?

A [Bob Supnik/Benn Schreiber]: Yes.

Q: What is the timeframe?

A: I can't comment on timeframes.


Q: I need VAX Pascal.  Will you port that to Alpha/NT?

A: We will consider it.  It should be easy to do once we have the
Alpha/NT backend working.  We have some of the languages product
managers here [I think he mean Cathie Richardson and Richard
Kaufmann] and you should make your requirements known to them.  We
will change our plans as we see opportunities.


Q: Will your other VMS languages be ported to NT, both for Alpha and
the other NT implementations (Intel and MIPS)?

A: We will look at this possibility based on opportunities, but
because of the GEM technology we will probably do this. [I should be
careful here:  I don't think Supnik made any commitments for us.  I
think he was very careful in his answers, even if he was a bit
optimistic on how easy it will be to get GEM moved around...]

Return-Path: tle::conti
Date: Wed, 8 Jul 92 10:39:29 -0400
From: tle::conti (Bob Conti 603 881 2086 Just DO It  08-Jul-1992 1033)
To: will
Subject: TLG mgmt team may enjoy this 1-of-N reports by Richard Wells
who's at the Win32 Developer's conference... Supnik on Alpha-NT languages!

From:	DECWRL::"[email protected]" "Richard A. Wells" 
8-JUL-1992 10:14:56.11
To:	me-dec <[email protected]>, bob <tle::conti>, Dieter
<4gl::heinzer>, nancy-b <tle::beckley>, bruce <tle::foster>, kathleen
<aimhi::bailey>, tom <tuner::bailey>
CC:	
Subj:	 

Question and Answers in Main and Selected Breakout Sessions, Day Two
Microsoft Win32 Professional Developer's Conference
7 July 1992, Moscone Center, San Francisco

Reported by Richard A. Wells (TLE::WELLS)

[For all of these talks the slides are in the proceedings which I
will be bringing back with me.  I have only paratranscribed the
question and answer sessions from each of these talks.]

In this issue:
  Windows Kernel [empty]
  Windows Executive
  I/O Subsystems [empty]
  Networking
  Win32 SDK
  Writing Distributed Application with RPC
  Writing Multithreaded Applications
  Alpha

Windows Kernel
--------------

[Cutler overran his time by 15 minutes (isn't he always on schedule?
:-), so there was no public Q&A.]


Windows Executive
-----------------

Q: [Edward Cheng, DEC] Is NT designed to support mission-critical,
high-performance features: TP mgmt, logging, etc.?

A: Yes.  We did this.  One thing we did in NTFS by making sure the
design could be extended to support transaction mechanisms.  


Q: You have logging in the executive.  Do you intend to support Open
Journal mechanisms?

A: Yes.  We should talk later.


Q: I'm very interested in doing Win32 UI calls and POSIX non-UI calls
in the same process.

A: The NT executive calls will not be POSIX compliant.  The POSIX
subsystem will be.  Lots of problems:  signal semantics, C run-time
differences. 


Q: Does MS have intention to document executive API to allow other
developers to build emulation subsystems.

A: Not in V1.  We want to retain the right to change this.  Win32
API, however, will be supported.


Q: Intend to do this in the future.

A: Maybe.  Talk to us.


Q: Can a [something about not exporting a processes(?) security
something-or-other.] 

A: Yes, through an anonymous LPC connection.


Q: Can you run HPFS, NTFS and FAT simultaneously.

A: Yes, obviously on different partitions.  Also a CD-ROM file
system. 


Q: What is going to happen in 4 years beyond 2GB address space.

A: We think 32-bits will be adequate for 5-8 years, at which time the
processors will support 64-bit addresses, at which time we will have
to figure out how to migrate 32-bit apps to 64-bit.


Q: Would it be good to put page files on physically separate volumes
from user files?

A: Yes.  The less disk reading the better.  If you can put page files
on volumes that are less often used will improve your performance, as
will having multiple paging volumes in heavily multiprocessing
environments.


Q: [Something about NT executive API.]

A: None of the emulation subsystems uses all of the NT executive
APIs, but the full API is available to all.


Q: Are Run-time libraries sensitive to the different emulation
subsystems.

A: Yes, there are separate libraries.  There are different call
semantics. 


Q: Is there going to be an X Window System available for the POSIX
subsystem?

A: We are working with third-parties.


Q: Can you query the granularity of thread quanta?

A: Finest you can set is in the milliseconds.


Q: On Intel platforms what support is there for building descriptor
tables (LDTs and GDTs)?

A: The NT executive API has calls that accepts LDTs.  If we like it,
we'll accept it, otherwise we'll return an error.  This is not
exported.  Only people that really need to do this is the DOS
subsystem. 


Q: Do you support running all sorts of DOS extenders under NT?

A: In the DOS emulation mode there is DPMI 0.9 capability.  We don't
support loading just any DOS extender.  Attend the breakout session
on DOS emulation.


I/O Subsystems
--------------

[Speaker ran over, so there were no public questions.]


Networking
----------

Q: How do we get a hold of the Winsock and streams specs?

A: Winsock is on CompuServe and the Internet.  Alistair Banks is our
connectivity developer relations person.  As for the streams spec,
you need to talk to our internal relations folks.


Q: [Question about POSIX sockets.]

A: We might not have sockets in POSIX release one, but we are trying
to. 


Q: [Question about access to Winsock from WOW.]

A: In this release we have only the synchronous parts of Winsock.
Async and mouse will be in the next release [I think he meant next
PDK update].


Q: 

A: We do not have generalizable directory lookup service
capabilities. 


Q: Will there be any mechanism for supporting SNA protocol stack.

A: There is the SNA services product.

Q: My question really is "will NT peers be able to communicate over
SNA protocol stacks"?

A: No.

Q: Never?!?

A: It could be done, but we don't have the resources to do all
transports.


Q: [Question about DLC.]

A: DLC drivers are not in today's kit.  They should be in the next
PDK.  They should also be in the SNA services beta, which should be
available soon.


Q: By using Windows NT do I get everything I need to set up a network?

A: Yes.


Q: When will we be able to use NFS to hook up NFS Unix system to NT
systems?  And what about FTP or TELNET servers for NT?

A: We will not do this in version 1.  There are third-parties working
on this, but they don't want me to tell you who they are.  Contact
our developer relations, who have more information.


Q: [SNA, mainframes]

A: [Third-parties]


Q: What is WOW?

A: Windows on Windows environment.  How you run Win16 apps on Windows
NT.


Q: Since there is now a global directory service, will it
interoperate with DCE.

A: We don't provide it, but a third-party could build it.


Q: Could your client directory service access DCE global directories?
Wouldn't you get it automatically since you have DCE naming support

A: You'll have to ask [person doing RPC talk].


Q: What about the DCE DFS?

A: No plans.  Licensing costs don't fit our market.


Win32 SDK
---------

[Demo]

All tools show up in a single program group (similar to the Win3.x
versions).  Contains tools and help files (cross-references into the
API, full descriptions of example programs).

Is going to build bitblt sample app.  Brings up command line prompt.
Goes to source directory.  Uses new NMAKE to compile it.  While it is
compiling, it changes context and demonstrates that the CompuServe
Information Manager to connect to CompuServe (actually dials up).

New LINK produces warnings for unused libraries that are specified in
LINK command.

Uses CIM to go to the WINNT forum.

Runs the program from the comand line.  Loads a bitmap file.  Selects
a region, copies it to another area in the client area, sets
parameters that demonstrate blitting (transforming) of bitmap.

Starts up Windbg (derived from QCWin debugger) to show running app
from debugger.  Sets some breakpoints.  Single steps.  Restarts app,
back to line 1 in debugger.

Goes to third-party tools.  Starts up Hamilton C Shell, which
provides most C shell functions, ls, more, diff (colored!) in a
command-line window.

System hangs.  "This is an opportunity for you to see Windows NT rebooting."


Q: One thing I noticed is that some of us like looking at something
other than black on white.  This is a problem in WinHelp.

A: I'll talk to the WinHelp devos.


Q: Will CompuServe be free for people here?

A: No, you will be charged for your connect time.


Q: Do the C++ classes have support for new Win32 features?

A: No, not in this first PDK.  More information later from guy giving
MFC talk.


Q: When we update to C7, definition of time broke, breaking our
application.  We researched this and found it was supposed to be for
ANSI compatibility, when apparently this doesn't even exist.  There
IS a POSIX standard of 1970 as the base.  Will you use it?

A: I don't know about this issue.  I'll put you in touch with C
folks. 


Q: Our product augments the resource compiler.  Is there facility in
NT for replacing components of the resource compiler?

A: [Didn't understand relationship to question...]


Q: When a program crashes, will GDI objects for that process be de-allocated?

A: Yes.  These resources are stored on the server side and they will
be freed when the client dies.


Q: Will second debugging machine need to be running NT?

A: We are looking at doing this in Win32s for Win3.1, but we aren't
sure. 


Q: Are you planning to come out with IDE?  In what time frame?

A: Not in this SDK.  The languages group will do this sometime.  They
will be talking about this tomorrow.


Q: Is new GDI in Win32s?

A: Beziers and splines won't be.  There is a talk tomorrow that goes
into the details of Win32s.


Q: Do you support WINDOWSX.H?

A: Yes.


Q: One of my problems is that the only video driver is VGA.  Help?

A: Only supports VGA and only some SuperVGA.  In the near future this
list will really expand.


Q: If I were to load this today and still kept my current MSC 7.0,
can I use it under NT to build 3.1 apps?

A: Not today.  We will be working in the coming months to do this.
Our focus today is for Win32 apps.


Q: Schedule for release of Win32s?

A: Probably in next update, hopefully September, but schedule is not
solid. 


Writing Distributed Application with RPC
----------------------------------------

Q: You mentioned VMS as a server?

A: Supported by DEC.  Shipped 1.0 starter kit (beta) about a month
ago for both VMS and ULTRIX.


Q: Mac client?

A: Seems like a good idea, but it isn't on our development schedule.


Q: If you start a server and then start app and app does RPC to abc()
and in the middle of that some other client starts up and calls
abc(), what happens?  One thread, two?

A: As long as you specify enough threads when declaring server enough
threads will be started up.

Q: Can you set it up so one thread handles them by queuing?

A: If you only allow one thread it will sort of work, but the
requests could time out.


Q: Could you shed some light on plusses or minuses between using RPC
vs. NetDDE?

A: NetDDE is more interpretive.  It works to applications that
support DDE now.  NetDDE won't perform as well but it would probably
be more flexible, due to its interpretive nature.


Q: What restrictions are there on the parameters, say there are
imbedded pointers in a structure, like a linked list?

A: You can have linked structures, as long as there no "aliased"
pointers.  The IDL expressions describe the structures.  As long as
there are not multiple pointers to the same thing (i.e. no cycles).


Q: If you have a lib that has a lot of static data, how do you handle
that?

A: Write routines that expect the static data and those call server
procs that take those as arguments.  Sometimes you have to have
wrappers around RPC stubs that turn things into constructs that RPC
can handle.  The jargon for this is "client agent" or "client clerk".


Q: [Missed it.]

A: Both sides of RPC use threading to get parallelism.


Q: What is load of RPC call?

A: Per message overhead is 16 bytes.


Q: Is there any way to automatically get load balancing?

A: Not yet.  This would be the perfect user-written server.


Q: How would you do this with C++?

A: No problem. You'd have to slap an extern "C" around the export of
the C header file.


Q: When will Win3.1 have client and server?

A: Right now there are no development plans for this currently.

Q: What about Win32s?

A: Right now there is nothing done, but it should be no problem.

Q: [How do you do a Windows server?]

A: Some experimental work we've done is a program with a hidden
window that gets its messages through its WinProc.

Q: In your example your server had to sit on this request loop.

A: Under 3.1 this doesn't make any sense, as you imply.  When we do
have a mechanism to have a Windows even delivered this will make
sense. 


Q: What facilities does the OS provide for continuation from crashed
clients or servers?  What cleanup will be done?  De-allocation, etc.
Are there exceptions that need to be handled?

A: If you are trying to right a long-lifed server, that needs to be
robust, etc. then you must be careful to catch exceptions and clean
up appropriately.


Q: When will there be OSF CDS and Kerberos components available?

A: There is a problem with publicly releasing anything that is
OSF-compliant, because they have not published royalty-free
specifications.  Only route is to license the source-code, which we
don't want to do.


Q: In your try/except block, what kinds of exceptions that come up on
the server will be made known to the client?

A: Yes, all exceptions on the server are reflected to the client.

Q: [Missed the question.]

A: LPC is an order of magnitude faster than RPC.


Q: I'm interested in doing load-balancing, with devos using Win3.1
systems, with an NT compute-server with automatic failover to the
local machine.  Is RPC the right way to do this?

A: It's the best thing I can think of for what you want to do.  You'd
have to write two servers:  an availability server and a compile
server.  The compile server would indicate its availability to the
availability server, and the clients would check there first, failing
over to a local compile.


Q: Data conversion:  how are you chasing down the user-defined types?
Where do the type definitions have to be for the MIDL compiler to see
them?

A: They have to be in the IDL file.

Q: I ask because CASE tools generate .H files.

A: IDL has #include statement [???]


Q: Can you pass pointers to unknown datatypes?

A: No.  We have local joke: if you want to pass a void*, we'll copy
your entire address space.  But you can specify uninterpreted data by
a size (basically a char[N]).


Q: Do RPC calls use persistent or non-persistent connections?  Can
this be controlled?

A: It uses connections that try to remain around as long as possible,
but are taken down if needed, like if the server gets overloaded.


Q: Can I fiddle with the packet after on the other sides of the
stubs, for example if I want to do compression or encryption?

A: No, goes against the philosophy.  There are mechanisms in RPC for
standard encryptions.  Compression you would have to handle at the
argument level.


Q: What is the relationship on connections?  Machine to machine or
application to application?

A: App process to app process.

Q: When is a server judged as saturation and when do connection start
being closed?

A: When a server attempts to open connections and can't.  We haven't
settled on a policy for what we close, although we try to go for the
oldest, least used.


Q: If I have a Windows client talking to an NT server, how is
security done?

A: Currently there is password exchange only under named pipes.  So,
this would have come during LAN Manager login.


Q: Say there are 15 of your phone book servers [sample code in
slides], does the client have any way of "browsing" them?

A: We are looking at this.  There may be something we can get into
the September release.


Q: In the network encryption routines, do you support complex data arrays?

A: Encryption handles anything, independent of the complexity of the
data types.


Q: Is the interface for local RPC?  How do you identify that you want
to do local calls?

A: You can encode it in the protocol sequence type.  Any server that
uses UseAllProtocolTypes will automatically pick up local transport
requests. 


Writing Multithreaded Applications
----------------------------------

Q: Can mutexes and semaphores be acquired recursively?

A: I think mutexes can.  I'm not sure about semaphores, but I think not.


Q: [???]

A: Within a given process all threads are co-equal, as opposed to
OS/2 where first thread was parent, and if it dies all others go
away. 


Q: Perf. data of thread switch overhead vs. context-switch overhead?

A: I don't know, I'm from the language group, not the kernel group.


Q: Does or will there be any C++ support for threads or
synchronization objects?

A: I'm not in on the C++ stuff.  I do know that the latest changes
we've added to the C RTL is a function-handler per thread.  I don't
think this is the PDK you got.  You'll have to ask the C++ folks.


Q: NT critical sections aren't like OS/2 1.x ones?

A: No, under OS/2 a critical section stopped all threads in that
process.  Under NT it is a synchronization data object.

Q: So it is data, not a section of code.

A: Correct.


Q: [???]

A: For shared memory you have a memory mapped file that has no
filename.  When one process writes to a block it will appear
immediately in the view on the file for the other process.


Q: Could I build a set of DLLs that each spawn sets of threads?

A: A DLL is just a block of code and data, so, yes, it sounds
reasonable.  Messages would need to be sent to the thread so that it
knows it is associated with that DLL.


Q: Is there a SendThreadMessage?

A: No, but a thread could put up an invisible window, poll the
Windows event queue and other threads could use Windows' SendMessage
to communicate with it.


Q: I'm curious about how cheap threads REALLY are.  How many threads
a uniprocessor is likely to be able to support?

A: Well, hundreds, at least, but of course it will depend on how much
physical memory you have and how fast your processor is.


Q: 

A: If you are talking about re-building a Win3.1 under Win32 you will
have conditional code.  If you use Win32s you will need to trap the
error message from thread creation (which you will get on Win3.1) and
behave differently in that case.


Q: Does the shared memory mechanism work over the network, perhaps
through NFS mechanisms?

A: I don't know.  You'll have to talk to NT or networking guys.


Q: [???]

A: You could have a thread wait on an event signaled by another
thread or on one from an I/O event and wait for any instead of wait
for all.


Q: Does NT provide tools for monitoring the execution status of
threads within a process.

A: There is an app in the PDK called performance monitor.  I don't
know how much detail it provides.


Q: If a thread has possession of a critical section and that thread
terminates, will the critical section be released?

A: I don't know.  That won't be true of a mutex, where the next
thread attempting to acquire it will get an error that the mutex has
been abandoned.


Q: Is there any intention to put the POSIX pthreads interface on top
of NT threads?

A: I don't know.


Alpha
-----

[I didn't have room to use my notebook (it was an SRO crowd,
apparently the most well-attended vendor presentation at the
conference (!)), so I'll just record some of the Q&A from memory.] 

General Comments:

  The presentation was well-received.  Supnik and Schreiber gave
  back-to-back presentations on the Alpha architecture and the Alpha
  program in general, and then on the NT port in particular.

  Key points:

  - Alpha PC will be made of STANDARD PC COMPONENTS.  This claim is
    real.  Supnik even showed a proto board.  The only hitch here is
    that there is a new internal bus called PCI, which I think stood
    for Processor Connect Interface.  We are promoting this as a
    standard with Compaq, IBM and others, which is probably
    sufficient to legitimize it as "standard".  Recently there has
    been a lot of activity in the high-end PC makers providing video
    boards that are in a very fast "processor-direct" slot, rather
    than on the EISA/ISA expansion bus (which is a relatively slow
    bus).  The PCI standard describes a very-high-speed bus with
    slots for high-performance options like video and SCSI
    controllers, as well as bridges to EISA/ISA buses.  The Alpha PC
    will have a PCI containing the video and a bridge to a six- or
    eight-slot EISA bus that will accept any existing PC expansion
    boards.  The NT design will allow drivers written for these
    devices for NT on Intel systems simply to be re-compiled on
    Alpha/NT.  All other components in the system are completely
    "low-tech" off-the-shelf PC parts.  They said this machine would
    be THE performance leader in PCs and would be
    competitively-priced with Intel systems.

  - The Alpha bits [:-)] for NT will be DISTRIBUTED ON MICROSOFT'S
    SINGLE, INTEGRATED Windows NT V1.0 CD-ROM.

  - First boot of Alpha/NT should be this month.


Q: Will DEC FORTRAN be ported to Alpha/NT?

A [Bob Supnik/Benn Schreiber]: Yes.

Q: What is the timeframe?

A: I can't comment on timeframes.


Q: I need VAX Pascal.  Will you port that to Alpha/NT?

A: We will consider it.  It should be easy to do once we have the
Alpha/NT backend working.  We have some of the languages product
managers here [I think he mean Cathie Richardson and Richard
Kaufmann] and you should make your requirements known to them.  We
will change our plans as we see opportunities.


Q: Will your other VMS languages be ported to NT, both for Alpha and
the other NT implementations (Intel and MIPS)?

A: We will look at this possibility based on opportunities, but
because of the GEM technology we will probably do this. [I should be
careful here:  I don't think Supnik made any commitments for us.  I
think he was very careful in his answers, even if he was a bit
optimistic on how easy it will be to get GEM moved around...]


Q: You must have some sense of performance compared to a 486.  What
can you tell us?

A: Well, it depends on your application, your instruction mix.  The
486 has fairly weak floating point and Alpha has very strong floating
point.  Not that you should compare systems this way, but if you want
to talk about SPECmarks, it is probably about 5 times as fast.

Q: Good. [Much laughter and applause.]




Distribution:
  me-dec >internet:[email protected]
  bob >internet:[email protected]
  Dieter >internet:[email protected]
  nancy-b >internet:[email protected]
  bruce >internet:[email protected]
  kathleen >internet:[email protected]
  tom >internet:[email protected]


% ====== Internet headers and postmarks (see DECWRL::GATEWAY.DOC) ======
% Received: by enet-gw.pa.dec.com; id AA03238; Wed, 8 Jul 92 07:12:45 -0700
% Received: by ihb.compuserve.com (5.65/5.910516)id AA02138; Wed, 8 Jul
92 10:12:40 -0400
% Date: 08 Jul 92 10:05:06 EDT
% From: "Richard A. Wells" <[email protected]>
% To: me-dec <[email protected]>, bob <tle::conti>, Dieter
<4gl::heinzer>, nancy-b <tle::beckley>, bruce <tle::foster>, kathleen
<aimhi::bailey>, tom <tuner::baile
% Subject: 
% Message-Id: <[email protected]>

re Note 1985.0 by RANGER::JCAMPBELL:

>     3. WNT boasts compatibility with Windows 3.x and MS-DOS applications.

        Binary compatibility is true only for WNT on the Intel
        architecture.

        This is why I wonder whether WNT on Alpha is that big a deal. 
        It will not run off-the-shelf software built for the Intel
        mass-market.  Even if it's just a matter of re-compilation,
        experience shows that the mass of software will not be
        available for a machine that isn't binary compatible.

        WNT may indeed turn out to be the best OS for the Alpha
        workstation, but does it really "catch the wave" or does it
        merely make a ripple?

        Bob

    re .9:
    
    Maybe we should have ANDF for NT.
    
    And I there might be Intel emulation for AlphaNT. Not an elegant
    solution, but at least it would allow to run Intel binaries with
    still reasonable speed.
    
    I cretainly agree with .0 that this is our great (last?) chance - and
    we have an advantage, DEC is full of VMS specialists, whose learning
    curve for WindowsNT should fast...
    

    The latest rumor to be making the rounds at Microsoft:
    
    From Arthur C. Clark's "Space Odessey 2001"
    
    	HAL + 1 = IBM   and, now from Dave Cutler  VMS + 1 = ???
    
    (The answer is left to the student as an exercise).
    

re :-.1

This was circulating on the usenet (and inside DEC a month or two ago).
Someone (@DEC) supposedly called Cutler and asked if it was true, and was
told "What took you so long?"

Chad

>                 <<< Note 1985.4 by FHOPAS::JAMBE::MCMULLEN >>>
>
>Re: .0   Small Nit - "Catch a Wave"  Jan & Dean

Nit reversal:  It WAS the Beach Boys who first did "Catch a Wave."
It was a surfing tune.  The Jan & Dean version had rewritten lyrics
as a SKATEBOARD version (new at the time, mid-60s) about "sidewalk surfing."

We now return to your regularly scheduled rathole.

<<< Note 1985.9 by LGP30::FLEISCHER "without vision the people perish (381-0899 ZKO3-2/T63)" >>>
                      -< I see two different things here >-

re Note 1985.0 by RANGER::JCAMPBELL:

>     3. WNT boasts compatibility with Windows 3.x and MS-DOS applications.

        Binary compatibility is true only for WNT on the Intel
        architecture.

        This is why I wonder whether WNT on Alpha is that big a deal. 
        It will not run off-the-shelf software built for the Intel
        mass-market.  Even if it's just a matter of re-compilation,
        experience shows that the mass of software will not be
        available for a machine that isn't binary compatible.

        WNT may indeed turn out to be the best OS for the Alpha
        workstation, but does it really "catch the wave" or does it
        merely make a ripple?

        Bob

==============================================================================

I was at the NT conference.  I don't have the handouts and notes handy to  refer
to, but the very strong impression I had was that there *would* be Intel
emulation on non-Intel NT platforms.  I had the impression that it would in fact
be required as part of the licensing agreement.  I believe that the MIPS
R4000-based NT that was running in the hands-on demo area already had it.  I
also seem to remember that Bob Supnik and/or Benn Schrieber said something about
their using Insignia Solutions(?) emulation software on NT/Alpha.

Two messages that I took from the conference (and scarcely the only ones) were:

1) Microsoft made a strong point on day 1 to say that part of their licensing
requirements were that every vendor would have to guarantee that NT would be
standard NT on every platform.  I take it that they have learned from the UNIX
fragmentation fiasco that has been so instrumental in preventing UNIX from
moving into so many of the markets that NT is aiming for.  I think that they
indeed have the right approach.

2) Microsoft's approach was that NT was of preeminent importance, over and above
what hardware platform it ran on.  (Naturally, for a software company!)  The
Intel emulation on non-Intel NT platforms is precisely to allow existing Windows
applications to be quickly available on those platforms, without the hardware
vendor having to convince every third party vendor to spend the money to test
their products and market them for every platform.  Without such existing
'shrink-wrapped' applications, it's clear that the hardware platform's appeal
for potential NT customers would be limited.

I am concerned that Digital's approach is to treat NT as merely another O/S (and
one deemphasized wrt our own proprietary O/Ss) with which to push Alpha hardware
-- and low-end Alpha hardware at that).  While the goal of selling Alpha
hardware is laudable, I think that we have it backwards;  NT is the important
thing, and we should be setting aggressive goals to port as many of Digital's
software products to *generic* NT as possible.  Not moving our software to NT,
limiting the number of software products being moved to NT, or moving to NT
Alpha and no other NT platforms are not roads to success. We are already seen
as a proprietary vendor (OpenVMS renaming notwithstanding), and this would
strengthen that image.  Not to mention limit our access to volume markets, which
surely we must move towards?

This conference convinced me that NT will be a large (even dominant) part of
the future of computing.  I hope the powers that be in Digital are doing the
right things so we don't miss this 'wave' like so many we have missed before.
I really do want Digital to succeed here...

Bryan

re Note 1985.14 by BROKE::HIGGS:

>               -< I think Intel emulation is in every NT system >-

        Software emulation is typically quite a bit less efficient
        than execution on the native architecture, i.e., for a given
        level of technology, the native processor should always beat
        an emulation by a sizable factor.  Yes, native Alpha will be
        significantly faster than the best "native" Intel processors
        available at the same time, but if for a given piece of
        software the Intel isn't emulating, and the Alpha is
        emulating, the Intel will outperform.

> Without such existing
> 'shrink-wrapped' applications, it's clear that the hardware platform's appeal
> for potential NT customers would be limited.
  
        Not only is the existence of off-the-shelf software on day 1
        important to the customer, but of almost equal importance is
        the anticipation of a stream of new, full-performance
        software in the future.  I don't think emulation is good
        enough to give this assurance, especially in the face of
        certain competition from systems that do not have to do
        emulation.

        Bob

re Note 1985.14 by BROKE::HIGGS:

>               -< I think Intel emulation is in every NT system >-

        This is really a rathole in the DIGITAL conference.

        I refer the interested reader to the DECWET::WINDOWS-NT
        conference, in particular topics 144 and 147.

        Bob

RE: .9

NT will have an Intel emulator on MIPS and Alpha.  It's important to
note two things about this:

1) Alpha is so fast that, even emulated, Intel software running on
   Alpha will outperform the same app on a native Intel system.

2) The emulator does not emulate calls to OS or Windows.  Those are
   done by thin jackets that immediately switch to running in native
   Alpha code.  Most Windows applications are known to spend the
   majority of their time in Windows rather than in application code.
   For such applications, the difference in speed between an emulated
   vs. native version of the app will be negligible.


RE: catching the NT wave

It certainly does represent an opportunity for us.  I know I'm doing
all I can to make sure we capitalize on it.

--PSW

    Re:  <<< Note 1985.17 by ADSERV::PW::WINALSKI "Careful with that VAX, Eugene" >>>

>1) Alpha is so fast that, even emulated, Intel software running on
>   Alpha will outperform the same app on a native Intel system.

Intel's P5 chip  (a.k.a.  586) is said to come at 100 MIPS.  I doubt that
a DOS emulation under Alpha will come close to native NT on a P5.

/OLD

re: <<< Note 1985.17 by ADSERV::PW::WINALSKI >>>
}
}NT will have an Intel emulator on MIPS and Alpha.  It's important to
}note two things about this:
}
}1) Alpha is so fast that, even emulated, Intel software running on
}   Alpha will outperform the same app on a native Intel system.
}

1. The NT Emulator extends only to Windows/DOS and not to OS/2.
   The OS/2 sub-system will not be available on RISC.

2. Even on Intel based hardware there is some emulation work carried out
   ... such as traps ... there is no need to have the Instruction Execution
   Unit which does Instruction emulation.

3. Intel (at the SFO conference) presented the P5 as a 100 MIPS
   processor and so I suggest it is not safe to assume that Alpha will
   still have such a dramatic performance advantage over the Intel base
   for emulated software

   Alpha will need (as I am sure will) win primarily on Native
   applications and other performance advantages.

- Steve.

RE:  <<< Note 1985.17 by ADSERV::PW::WINALSKI "Careful with that VAX, Eugene" >>>

>1) Alpha is so fast that, even emulated, Intel software running on
>   Alpha will outperform the same app on a native Intel system.

Emulation will be faster than some Intel systems (25 MHz 386?), but not faster
than the high end 486 or 586 native system. 

Dileep

There are also rumors that Intel engineers are having trouble even
approaching the performance that Intel marketeers are telling people P5
will have.

Sheesh!

While many of you are drooling over this leadership processor technology
between yourselves, there are a fair few customers with some magic words
stamped on their foreheads:

			"What's in it for me".

You won't catch any wave if you don't answer that one first, IMHO.

								- Ian W.

Article: 32091
Path: ryn.mro4.dec.com!nntpd.lkg.dec.com!news.crl.dec.com!deccrl!caen!zaphod.mps.ohio-state.edu!mips!mash
From: [email protected] (John Mashey)
Newsgroups: comp.arch
Subject: Re: What's in a name?
Date: 16 Jul 1992 22:31:04 GMT
Organization: MIPS Computer Systems, Inc.
Lines: 37
Message-ID: <[email protected]>
References: <[email protected]> <[email protected]> <[email protected]>
NNTP-Posting-Host: winchester.mips.com
 
In article <[email protected]> [email protected] (Craig Dowell) writes:
 
>There'll be a 686 all right.  It may look like the x86 architecture from the
>outside, but on the inside I'll bet that it'll be built so that it outperforms
>any competetive product.
 
Well, maybe; on the other hand, the 586 is rapidly running into reality;
- the Intel charts last Fall show it 4-5 months behind the R4000,
and at 70 SPECint.  For this to be true, it means 586 production *systems*
must be shipping no later than a few weeks from now, as SGI started
shipping R4000s in Crimsons no later than the beginning of March.
 
First, they were going to to announce very soon, and be able to speak at
Hot Chips.
They dropped out of Hot Chips.
Second, they're supposed to talk at Microprocessor Forum ... but that's
getting iffy, too ... but they'll be announcing at COMDEX...
Well, maybe....  Consider that they announced 50Mhz 486 around June 91,
and it was 6 months before buyable systems appeared, and that was
a small chip, not a monster like the 586, in BiCMOS, which doesn't
shrink like CMOS, and Intel doesn't generally do, using a die
40% bigger than an R4000....
 
Maybe the 686 will be better .... however, although what competitors
say about each other should *always* be taken with a grain of salt,
a *lot* of what Intel tells people about chip futures, how RISC-vs-CISC
doesn't matter when you have lots of transistors, etc, is BOGUS, and it's
bogus in a way that anybody who designs chips knows perfectly well,
and in fact, as is perfectly laid out in Mead & Conway from 10 years
back, Chapter 1.... which just about any undergraduate EE would have
studied.
 
-- 
-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: 	 [email protected] [soon to be [email protected], but not quite moved yet].
DDD:  	408-524-7015,  or 524-8253
USPS:	(soon) Silicon Graphics, 2011 N. Shoreline Blvd, Mountain View, CA 94043

The following is an attempt at a transcription of "Catch A Wave" by the Beach
Boys. I realized, when I replayed it, that it was astoundingly relevant
to the computer industry today. See if you see the same symbolism I do.

			Catch A Wave

						by Brian Wilson

Catch a wave and you're sitting on top of the world.

Don't be afraid to try the greatest sport around. (Catch a wave, catch a wave)
Those who don't just have to put it down.
You paddle out, turn around, and race in, baby.
That's all there is to the cosine craze. <---is this right? Funny!
You gotta catch a wave and you're sitting on top of the world.

Not just a fad 'cause it's been going on so long (catch a wave, catch a wave)
They said it wouldn't last too long.
They'll eat their words with a fork and spoon and watch 'em
They'll hit the road and all be surfing soon
And when they catch a wave they'll be sitting on top of the world

Catch a wave and you're sitting on top of the world.

So take a lesson from a top notch surfer boy. (catch a wave, catch a wave)
But don't you treat it like a toy.
Just get away from the shady turf,
and baby, go get some rays on the sunny surf,
and when you catch a wave you'll be sitting on top of the world.

Catch a wave and you're sitting on top of the world.

Catch a wave and you're sitting on top of the world.

    

>	That's all there is to the cosine craze. <---is this right? Funny!

	No, it isn't. It's *coastline*...

				Tom_K

    You might ask: why on earth have I started this note in the Digital
    notesfile? The answer: I think in order for this Corporation, which
    has some of the best hardware and software engineers in the world
    working for it, to become profitable again, we need to figure out
    how we can change the way we respond to new innovation in the
    computer industry.
    
    When IBM introduced the PC years ago, and Apple introduced the MAC,
    we talked about them as "toys". For many engineers here, they are
    still toys. Some of us have one at home, and it does our checkbook
    balances and our spreadsheets. But they are not important in our lives.
    We work on VAXstations, or RISC UNIX workstations, using a "real"
    operating system.
    
    But behind our backs, the major corporations got a different message
    when PCs began to be cheap and fast. Two things happened:
    
    1. A little company called Novell
    started connecting them in PC networks - Netware - which we referred
    to jokingly as "weeny-net".
    
    2. Little companies such as Borland and Lotus started writing
    high-quality software for these little toys, and sold it for next
    to nothing ($100-$500), because millions of customers times next-to-nothing
    equals a great deal of money. (Many of the ideas for this software
    came from Digital, but we didn't choose to make it run on the Intel
    platform. Lotus Notes and CCMAIL are good examples.)
    
    Today, over 70% of the "desktop devices"
    on people's desks are PCs, and some 10% are MACs. There have been 100
    *million* PCs sold, or about 100 times as large as the total VMS
    base.
    
    The corporate computer buyers got the message we missed: they could start
    putting together a desktop computer net that would eventually be the
    lifeblood of their company. Insurance companies and banks have begun
    the transition from their mainframes to running their business on
    Netware. (This is no joke. One of the largest insurance companies in
    the US does their stock portfolio trading on 386SX machines running to a
    Compaq Superpro server running Netware).
    
    With the advent of NT (which has a built-in Network Operating System
    called LANManager), and the high-end Intel chips, it is clear that
    Microsoft and Novell will continue to dominate the PC and PC/Server
    market. More and more companies will move their applications to the
    PC/file server environment, displacing mainframes, midis, and minis
    forever (except maybe as file servers, but even that is questionable
    considering that an Intel 100 MIP box would probably run $10,000 with
    more memory and disk than a mainframe comes with today...).
    
    Instead of buying from a salesperson, an MIS manager will buy by
    phone from Egghead Software or have the software pre-loaded when the PCs
    are purchased from Gateway. The existence of a direct sales force may
    be superfluous. The software that runs on the PCs is self-instructing,
    and the manuals that I have seen are excellent, far superior to most
    Digital documentation. There is a grand competition for management
    software, voice software, scan/image software, FAX software, database
    software, drawing/painting software, word processing and publishing
    software, tax preparation software, stock portfolio software. Each new
    entry is far better, and usually cheaper, than its predecessors.
    
    My question is: how can Digital meet this challenge and thrive as a
    company? What would it be like to have Digital competing against
    Gateway? What would it be like to have Digital software, in flashy
    shrink-wrap boxes, on the shelves at Egghead Software? What would it
    take to get us there?
    
    The Beach Boys were really computer industry prophets. There are
    waves of technology changes. If you catch one, you can be "sitting
    on top of the world" - until the next wave. Digital *made* some
    of the first waves - the "affordable" computer (by 1960's standards),
    timesharing, and workstations. We didn't catch the last wave.
    Will we catch the next one?
    
    							Regards
    							Jon

    Excellent Polemic...!
                      
    Next question, is, how true is it!
                      
    100 Million PCs vs Million VAXen (or is it VT100s connected to VAXen)
                      
    Remember you connect a fair few users to each VAXen - and each one can
    run MULTIPLE sessions using a LAN - on the PCs I tried, I couldn't even
    read mail at the same time as compiling Quick-C.             
                                                                 
    Yet nevertheless, the message is there. We are probably TOO late to
    beat MicroSoft or Lotus on PCs under MS-Windows, but NT is a different
    ball-game - and as networks appears to be analysts falvour of the
    decade, why shouldn't Digital, having the largest private peer-peer
    network in the world, not dominate the distributd computing business ?
    
    
    Winton                                                       
                                                                 
    p.s. If someone coughs up a 386 PC for me to use alongside my old
    VAX2000 W/S I'll give it a go...

    re: .27 Hi, Winton
    
    If you find a 386, try running PATHWORKS on a VAX somewhere and run
    eXcursion on your 386. It out-performs VS3100s running Windows.
    
    re: Catch a Wave: correction: the line with the question mark
    should be:
    
    "That's all there is to the coastline craze".
    
    I didn't think Brian Wilson was a mathemetician...
    
    Now, for more serious matters: how are we to catch the wave?
    I agree with Winton, we should be in a position to be in the lead
    of peer-to-peer networking, but Novell ate most of our lunch a
    few years ago with Netware, and Microsoft is about to eat the rest
    (along with Novell's), by realizing the potential and popularity
    of the Intel platform for doing real computing. While were thinking
    of PCs as toys, Microsoft, Novell, and several *thousand* other
    small firms saw the opportunity to write real software for them.
    
    To make the change to profitability, I think we need to bring
    ourselves into 1992:
    
    1. Get a PC (or MAC, or both) on everyone's desk, and turn the VAXes
    into file servers. That will bring us to the point that most U.S.
    Corporations are at. The PCs should be running Windows 3.1 or
    evalutation copies of NT, the MACs should be running MAC-OS V7, which has
    peer-to-peer file sharing.
    
    2. Get everyone to think in terms of *real* market share for our
    hardware and software: millions of copies, at a few hundred dollars
    (maximum) per copy.
    
    3. Port those products for which we have a technology advantage to
    NT *AS THEIR HIGHEST PRIORITY*. Sell them at Egghead in flashy
    packages, in Computer Shopper, and via third-party vendors. A direct
    sales force is an anachronism for this business. (The sales force
    is useful for the very large, complex contracts with the largest
    corporations.)
    
    4. Port our customer base to NT when and where it makes sense. Many
    of our classical VMS customers have been shifting their use of their
    VAXen to PATHWORKS file servers, shifting the real computing down onto
    the desktop. Those customers who have moved most or all of their
    applications to PCs can shift to NT (on Alpha or whatever architecture
    they want) very easily. VMS presents few real advantages over NT
    for most applications. Continue to support VMS on the VAX architecture
    and on alpha, for those customers who cannot easily port their
    software.
    
    Does this make sense?
    
    							Thanks
    							Jon Campbell
    

    I tell you what really annoys me  (politely!) is that the grunts can
    sit here and could be in a position to make the decisions, yet there
    are HIGH-LEVEL managements who are spending months (literally) trying
    to make up their minds what to do. 
    
    I *personally* would have thought it should be quite easy to decide on
    a strategy, and then do it. The trouble seems to be that the ones in
    power just can't - either no vision or no knowledge. 
                                                        
    Does anyone disagree with -.1 for example ? If not then make him the
    CEO.
     
     Winton

    For my money .28 is pretty near spot on. Some of us here have been
    *begging* for PCs for months, and months. Only this week, we had some
    more brand new VAXStations arrive... Almost without exception, these
    will be virtual terminals, no-one gets a 'real' Station.
    
    I promise you Winton, take a good long look at a 386, with PATHWORKS,
    eXcursion, etc. then take a good long look at the software these
    'toys' run, especially development environments, object-orientated
    *everything*, and then having asked yourself where the VAXStation fits
    in, you'll have more than enough food for thought.
    
    Laurie.

It is quite easy to "decide" what to do when your decision doesn't mean
anything (like in a notefile) but when you are really deiding the fate of
120,000+ individuals and the future value of investments for thousands of
others itis not so easy. Alternatives must be carefully analyzed, consequenses
planned for data checked and rechecked and then, in the middle of the night
you, who are being depended on by the multitudes to be brilliant, must make
a final decision with not much more to go on than your gut.

For example, why are elections polls so often wrong? Because people know that 
the election isn't being held "today" and their answer means nothing.

or 

Why are the decisions of the Supreme Court proving more middle of the road 
than the condservative makup of the court lead analysts to predict? Simple
the decisions being made by O'ConnerSouter and Thomas now count for far more 
than they ever did before.

                    <<< Note 1985.28 by RANGER::JCAMPBELL >>>

>    If you find a 386, try running PATHWORKS on a VAX somewhere and run
>    eXcursion on your 386. It out-performs VS3100s running Windows.
    
     I've got this configuration in my office and Jon is absolutely on the
    mark.  
        
    
>    To make the change to profitability, I think we need to bring
>    ourselves into 1992:
    
     I work for Jim Liu in the Personal Computer Group and we're very busy
    developing products that support your proposal.  
    
    >    1. Get a PC (or MAC, or both) on everyone's desk, and turn the VAXes
>    into file servers. That will bring us to the point that most U.S.
>    Corporations are at. The PCs should be running Windows 3.1 or
>    evalutation copies of NT, the MACs should be running MAC-OS V7, which has
>    peer-to-peer file sharing.
    
    Most, if not all of PC Engineering is configured with PCs attached to a
    VAX-based file serving network running Pathworks.      
    
    
>    2. Get everyone to think in terms of *real* market share for our
>    hardware and software: millions of copies, at a few hundred dollars
     (maximum) per copy.
    
    We operate by what we call the "commodity business model" and are
    developing our product plans based on high volumes (ie, hundreds of
    thousands of PCs over the life of the product.  As an example, at a
    recent product meeting, we shot down shipping a certain label that
    costs about 35 cents as we couldn't justify the cost over the life of
    the product. 
    
>    Does this make sense?
    
    To PCG it does.    
    
    Mark.    

    

Dear David,                             

    Below is a copy of a note that I just sent to Bob Palmer. I read in
a Wall St. Journal article that you and he will be charting a strategy
to bring our company back to financial health, so I thought I would share
my insights - which I believe are from a unique perspective in the company
- with you as well. It is going to be a great challenge and adventure,
and for the sake of this Corporation and its outstanding workforce I
wish you the best. I will be doing my part in PCSG to meet the challenge.

						Thanks
						Jon Campbell
						PATHWORKS for OpenVMS-Alpha
						Tech lead


Dear Bob,

    You don't know me - I'm the lead engineer porting PATHWORKS to run on
Alpha - but I wanted to congratulate you on your elevation to CEO, and to
provide you with what I think is a unique perspective on what Digital can
do to become profitable again.

    I'm in a unique position here in PCSG: I work with high-end PCs, and
write software for VAX (and now Alpha) to allow PCs to connect to them
as file servers. I see what our largest customers have (which is a PC
on every desktop) and what they need (good, fast file servers, top-notch
software running on the PCs that utilize their interconnection in a seamless
manner).

    Below is a note which I put into the DIGITAL notesfile, on node HUMANE,
which talks about what I believe to be the next revolutionary phase of
software and operating systems, and how it might affect our Corporation.

    In short, I see the computer industry technological advances as a series
of waves. In the past, we caught or even created those waves - low-cost
modules, minicomputers, time-sharing. But the last two - the advent of the
PC and the networking operating systems connecting them - we missed, perhaps
because we didn't see them coming.

   I see at least one wave coming: Windows-NT from Microsoft.
I believe that, along with the innovations, price reduction, and
speed-up of Intel & Digital hardware, and the proliferation of
PC software, that this wave will dramatically change the course
of computer usage.

    Many of the Fortune-1000 companies are now running, or contemplating
running, their entire businesses on PC networks. Windows-NT
is a real, virtual-memory operating system that has many or most of
the features of VMS and UNIX. It runs all MS-DOS and Windows applications
already in place. This makes it the vehicle by which the traditional
mainframes and mid-range computers will be displaced by file servers,
with Windows-NT running on the desktops (where necessary for virtual memory
applications) and Windows-NT running on a high-end Intel or Digital file
server.

     The challenge, I think, is for us to find the way to catch this wave
this time, and to catch the potential waves on the horizon (e.g., Newton
from Apple), thrive, and take the lead again where we can.

						Thanks for your time
						Jon Campbell
    
    (attached was a copy of note 1985.0)

If we want to catch a wave, we should be selling our networking expertise to
get in on the business of connecting PCs up to the Internet, along the
lines of what Compuserve, MCI etc are doing. It's obviously a profitable
business, otherwise they wouldn't be doing it, and you can see just how
profitable it is by examining the annual accounts of these companies.
Even in the UK, there are over 8 companies offering Internet and
Bidirectional usenet access for at little as �10 a month. Even IBM's
in on the act. It would be a real reversal of times past if Digital
PCs were being connected up via an IBM network.

DECWRL:: has been one of the world's most important routing nodes
for years, yet we seem reluctant to move into the market of personal
networking. If we hang around much longer, the companies starting up now
will become too established in the marketplace. Surely Digital, the
world's leader in networking can do something to connect up the
millions of PCs out there to real networks ?

I proposed this idea in the UK Digital conference, and some people
thought that this could be a big money earner for Digital.

Since then, I have compiled a listing of what services are on offer
in the UK (topic 115 in the GB92 conference) and suggested the idea
to Delta. If anyone thinks this is a good idea to pursue, then please
comment here and suggest where I can take it from here.

Craig

    <Windows-NT is a real, virtual-memory operating system that has many or
    <most of the features of VMS and UNIX. It runs all MS-DOS and Windows
    <applications already in place. This makes it the vehicle by which the
    <traditional mainframes and mid-range computers will be displaced by
    <file servers, with Windows-NT running on the desktops (where necessary
    <for virtual memory applications) and Windows-NT running on a high-end
    <Intel or Digital file server.
    
    And it is being built by the same man who gave us RSX-11M, and	VMS
    which not accidentally is 1 letter earlier than			WNT

    (Dave Cutler)

    For a good perspective on this note string, read _Accidental Empires_,
    by Robert X. Cringeley.  "Cringeley" is _InfoWorld_'s gossip columnist. 
    While a lot of the book is gossipy and superficial, some of Cringeley's
    points speak directly to the opportunity discussed here.  I'd like to
    mention three of them -- a spur to read the book.
    
    1.  Microsoft is not a company devoted to technical excellence, high
    performance products or even good product quality.  It simply wants to
    be ubiquitous.  Windows NT is hardly a bundle of exciting, new
    technological developments, but Microsoft will push hard to make it
    ubiquitous.
    
    An aside:  In 1987, I bought a personal computer first introduced two
    years earlier.  It featured then both command line interface and
    windowing, plus preemptive multitasking, shared reentrant libraries, an
    object-oriented OS kernel, almost perfect compatibility up and down the
    product line, plus graphics accelerator and audio output standard.  It
    still does, and most of my 1987 software still runs fine on the 1992
    operating system version.  _Byte_ magazine has carried many technical
    details on these poing. Today, outside its video niche, with an
    installed base of 3 million units, it's considered a failure.  My
    point: technical excellence has little to do with market success if
    you're not first or second in the field.
    
    2.  Digital, like all middle-aged computer companies, is concerned by
    and large with protecting its established business.  Breakthrough
    products, and wild success, depend on ignoring the past.
    
    3.  People buy new kinds of computers when there is a "compelling
    application" to justify the purchase.  In that case, other factors are
    irrelevant.  Three such applications are VisiCalc, Lotus 1-2-3 and
    Aldus PageMaker.  It's worth noting that the first and third broke new
    ground for individuals.  1-2-3 was compelling because of its
    performance, gotten at the expense of portability.  Point: strong
    success in software depends on adapting to the best features of the
    machine.  Mere "porting" does not create standout applications.
    
    Wes

let me interject some cold water onto the surfing note ....

To "catch the wave" we need the tool (the surfboard or PC in this case). To
order and receive a PC internal to DEC is just about impossible to do in any
reasonable amount of time (reasonable is less than three (3) months). I buy for
VMS Development and we have been honestly trying to switch the desktop platform
to PC's from x-terms and workstations so that the engineers begin to see how
the real world hooks back into VMS platforms. 

The internal supply chain for PC's does not work. We will miss this wave 
because our own internal process won't let us ship a PC inside, no PC's
inside means no development of product, no development of product means nothing
to offer for sale, and no sale means no chance for profit.

I have orders outstanding for 44+ PC units. Some of the orders are 8 months 
old. You don't go outside because that is not cost effective. So as far as
catching the wave ... not a chance ... so we don't have the tools so we will
sit on the beach and be "surfer wanabees".

----------------------------
Now this picture can change IF and only IF manufacturing can make enough
of the PC's available to IEG to satisfy some of the internal demand. Don
Quixote would like that kind of challenge.

Regards,
/jon

    I agree that it is ridiculous that you can't get PCs internally from
    Digital. If the ordinary channels don't work, you might try calling
    1-800-DIGITAL and order as if you're calling as a real customer.
    
    If that doesn't work, you might consider buying from Gateway or Dell,
    and it's probably a good idea to have some of these units anyhow
    because we want to be sure that the software we create works on the
    most popular vendors' platforms...
    
    							Jon

    
    
    
    	How many VP signatures would you need to obtain to get one of 
    	those ?

Strangely enough I can get one/two Dell/Compacq/other vendor PC with but one
signature besides mine. But that is not cost effective. For those D.E.C. has to
pay real money. Having one or three or so  as test platforms is ok. Buying 100+
isn't. 

I'd order the PC's on my credit card through PCBYDEC if I thought I could get
the reimbursment through the financial audit questions that a $35,000 credit
card bill would raise. And having been on the receiving end of many audits, I
know what chance me and the snowball would have.

/jon


Note: IEG money is also very real. But it is much cheaper to spend IEG money
	than external money, and especially for a product that only has a 
	2-3 year life span.

Buy an IBM PC, you can get those delivered quick enough...

    I'm not sure how much money we save by buing internally!
    
    PCs are a low-margin business.  Our traditional 27% IEG pricing makes
    no sense there; if Jim Liu were to let PCs go via IEG at 27%, then his
    business unit (set up to run at very low margin, high volume) would be
    losing lots of money on every sale.  That's not like traditional VAXes,
    where the 27% usually covered manufacturing's transfer.  (This, of
    course, is simply a reflection on what's happening to the industry, and
    to our margins in particular.)
    
    Our actual cost of PCs is probably competitive with other like vendors'
    costs.  Some of them are charging very, very small markups.  If the PC
    you need is NOT quite as fancy as Digital's own (and we tend to make
    "better" rather than "cheaper"), then the true cost to Digital of
    buying outside may be lower than buying inside.
    
    You can buy a nice 386-25/4M/SVGA/80MB PC for $1100 from a local store. 
    It won't perform quite as nicely as our own, but if that's what it
    takes to get people "up to speed" on what our customers are using, then
    it looks like a good deal to me.  Heck, get 'em to put the monitor on a
    separate line item and it'll probably be "expense", not "capital"!

    Quick question, surely a PC is a PC ? Bit naive I guesss, but doesn't
    everyone use the same basic chip-sets,drives etc ? I know most have
    different packaging - but what makes the difference  - assuming were
    talking equals 486DX v 486DX ?
                                  
    Winton

    re: .43  (WDAVIES)
    
>    Quick question, surely a PC is a PC ? Bit naive I guesss, but doesn't
>    everyone use the same basic chip-sets,drives etc ? I know most have
>    different packaging - but what makes the difference  - assuming were
>    talking equals 486DX v 486DX ?
                                  
    It turns out that not every motherboard design with a yMHZ x86 chip on it
    performs the same as every other design.  If you look at computer
    reviews in _PC_Magazine_ you usually see Compaq machines outperforming
    J. Random Clone boxes.  The BIOS you pick, the support chipset, and
    especially things like cache (speed, amount, etc.) have a major impact
    on usable speed.
    
    I have no idea how many companies there are out there who design
    motherboards, but it must take quite a bit of engineering.  Why DEC
    hasn't gone into this area (designing and manufacturing higher
    performance but non-proprietary x86 motherboards) is unknown to me.  It
    certainly seems like something we have the expertise/resources/need to
    do and could capitalize on.  (If we already do it, I haven't heard
    about it and we sure don't advertise the fact.)
    
    -joe tomkowitz

    re .43
    
    "PC" means personal computer to me.  It means any computer that I can
    use "personally".  That would include an Apple, an Amiga, or and IBM
    clone.  I know that's not the "PC (politically correct)" attitude, so
    sue me!  :-)
    
    Dick

re: .46

>    (heard about a thousand times) "When is this going to be available
>    under MS Windows?" or "Do you have a client/server version of this

What is "this"?  I realize there may be several answers.  So I'm really asking,
what are the products that we have that your customers would like to see 
ported to Windows, Client Server, or IBM/SNA environments?

   Gary

    re .44
    
>    I have no idea how many companies there are out there who design
>    motherboards, but it must take quite a bit of engineering.  Why DEC
>    hasn't gone into this area (designing and manufacturing higher
>    performance but non-proprietary x86 motherboards) is unknown to me.  It
    
    The reason we don't do it is because we now have the best in the
    business doing it for us - Intel.  Tandy and Olivetti built our earlier
    and lower-end machines, but the new 400 series have Intel motherboards,
    and right now are the only boards on the market designed for i586
    upgrades.

    
>    I have no idea how many companies there are out there who design
>    motherboards, but it must take quite a bit of engineering.

	It used to, but with the advent of off-the-shelf ASICs and VLSI
	chipsets, it's not that major an undertaking any more.

	I just bought a no-name clone 386SX-25 motherboard. It has a whopping
	19 chips on it (not counting memory, of course)! Only 5 of the chips
	(CPU, keyboard controller, CMOS memory, BIOS, and "chipset" (only
	one chip, actually)) could be considered LSI density or above.

Larry

    We should and are building Pathworks for NT over VMS and ULTRIX and
    OSF/1 and pray we get some of the disk server business.
    
    As far as selling desktops with NT, it'll be difficult to match INTEL's
    market lead advantages.  Like answering the question of INTEL emulation 
    on non-INTEL platforms.  
    
    karl

    
    RE: Intel being the best in the business. 
    
    They are brand new to the motherboard business and having worked 
    with them now for over a year, they are still learning. They also
    don't come cheap. 
    
    The all DEC-built and designed Tiger PCs  are what we need. They
    are currently for sale in GIA, but not Europe and US. They are
    both better and cheaper than the Intel and Tandy supplied PCs
    and we need to replace those vendors with teh Tiger products
    asap. 
    
    Jim Lui is on the right track. 

    re: .46
    
    There are a dozen or two Digital software products that are
    "best-in-class" that we might choose to port to NT, and in my opinion
    we would do well to port them as our highest priority.
    
    I'm not familiar with all of them, but here are a few:
    
    1. RDB and its associated applications.
    
    2. LSE. (Yes, that's right. Very few if any of the development
    environments come with good language sensitive editors...)
    
    3. Some of our expert (AI) systems.
    
    4. Our TP products.
    
    5. The GEM back-end and our language products (this is being done).
    
    6. DEC's super-accurate math runtime library (I'm not kidding about
    this. Mary Paine's work on our math library has made it the most
    accurate in the industry; it is, literally, "good to the last bit."
    
    7. eXcursion - the X-windows interface for MS-Windows (this is being
    done).
    
    8. Bookreader
    
    9. Remote System Manager (RSM) -  Large enterprises need a mechanism for
    loading application and system software and upgrades onto hundreds or
    thousands of PCs in an orderly fashion. Our current version, which
    remotely loads software and operating systems onto VMS or ULTRIX
    clients and provides automated, scheduled backup, is
    somewhat cumbersome, but the functions it provides are unique.
    
    10. The VMS debugger. There are few, if any, debuggers that even
    come close.
    
    I don't know the VMS layered product set enough to know which others
    are best-in-class (or, like RSM, are unique).
                                              

>    2. LSE. (Yes, that's right. Very few if any of the development
>    environments come with good language sensitive editors...)
     Does anyone knows how many of major software companies out there uses
     such editors (language sensitive)? 
     I think Emacs is most popular 'out there' , I dont know much about 
     Emacs, but I think it is a programmable (macros) and very versatile.

>    6. DEC's super-accurate math runtime library (I'm not kidding about
>    this. Mary Paine's work on our math library has made it the most
>    accurate in the industry; it is, literally, "good to the last bit."
     Most of this is written in macro , I believe, and customized to 
     VAX/ALPHA hardware, how easy would porting this be?
        
>    10. The VMS debugger. There are few, if any, debuggers that even
>    come close.
    I think the future fashion in an object-oriented debuggers, where you can
    examine an 'object' etc... Microsoft was working on such debugger few
    years ago. dont know current state.
    yes, VMS symbolic debuger is good, but also is a window based debugger
    I used 3 years ago on SUN workstation, that was neat debugger.
    
    /nasser
    ps. you forgot to mention Xdelta. ;-)  
    

    re .50
    
>    both better and cheaper than the Intel and Tandy supplied PCs
>    and we need to replace those vendors with teh Tiger products
    
    How do you substantiate this?  We haven't releasd the higher-end Tiger
    PCs yet, and our existing high-end - which are Intel - are
    outperforming competition like Compaq (I've seen the comparitive
    benchmarks).

    re .51
    
>    10. The VMS debugger. There are few, if any, debuggers that even
>    come close.
    
    Actually, I've been using the debugger that comes with Borland C++.  At
    the level I'm using it, at least, it compares quite well with the VMS
    debugger.  In fact, I like it better than our DECwindows/Motif version
    of the VMS debugger...
    
    /Dave

        <<< Note 1985.53 by UTROP1::SIMPSON_D "just call me Lazarus" >>>

    re .50
    
>>    both better and cheaper than the Intel and Tandy supplied PCs
>>    and we need to replace those vendors with teh Tiger products
>    
>    How do you substantiate this?  We haven't releasd the higher-end Tiger
>    PCs yet, and our existing high-end - which are Intel - are
>    outperforming competition like Compaq (I've seen the comparitive
>    benchmarks).
    
    .50 is correct.
    
    Intel makes our DECpc 433 Workstation and Tandy, of course, makes our
    current portfolio of desktop systems.  I believe you're referring to
    the 400st series.
    
    Mark.

    
    RE: Intel/Tandy vs Tiger solutions. 
    
    You are right, we haven't released the high-end Tiger. But I'm
    very familiar with the product. The 433 version is higher 
    performance than the PC433W at almost half the transfer cost. 
    
    In the low-end, the DECpc 320SXL
    (the Tiger version) is selling for $965, but only in GIA. Compare
    it with the DECstation 320SX currently selling through DECDirect 
    for $1,139. The $1,139 is actually a good price, competitive with
    DELL. TIGER would be all alone at $965!! We have to get Tiger into
    the US and Europe quickly. 
    
    The 400ST is a great performing product, no question. It's even selling
    well. We just aren't making that much on it. We have to keep that
    in mind, and not just focus on raw performance numbers. 
    It's just too expensive reselling PCs from other people. The margin
    pressure doesn't allow it. Tiger is the answer to that.
    
    So my point about Tiger wasn't that we have it now. We have to 
    get it soon. 
    

re: .51 and follow-ons

There's a fundamental flaw in the traditional analysis of tools that
frequently causes us to overrate ourselves.  That flaw is the
assumption that neat features make the product better.  The correct
principle is that neat features that will be used make the product better.  

Both LSE and VAX Debug are feature laden.  The new DEC Debug GUI, which
I gather is being well received, presents just a small subset of
available features on VMS (with the others available on the command
line).  In an old survey of LSE users, we learned that the
compile/review cycle was highly used, but template expansion was less
frequently used (and I sometimes think that the only people to use
LSE's overview capability all work in ZK2).  LSE beats emacs in ease of
customization of templates, and a few other areas, but not in the out
of the box template mechanism; emacs beats LSE (and most other editors)
in its indefinite undo, wide range of packages, and, of course, price. 
I'm just starting to explore PC program development editors, and I'm
prepared to be impressed.  There are some world-class ideas and
features in LSE, but overall, I don't think I would put it there. 
(Speed it up, give it indefinite undo, recursive edit, and pure emacs
emulation, then it will be the best.)

As for bookreader, I understand that a new version, which may be world
class, is on the verge of entering field test.  The version that I'm
currently using compares poorly to the help tool on MS Windows.

   Gary

A few replies back I'd said throw "cold water" on your wave ---

Now let me park the ICEBERG off shore.... Just got a mail from IEG telling
me in essence that any PC order I had would be returned, any new order would
be returned and try again in January. Returned as in cancelled.

Unless there is some radical change in the PC space, the closest we will get to
this wave is listening to the radio and reading the magazines. Much like an 
Eskimo did when the surfing craze hit southern Cal.

/jon