| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 769.1 | | TLE::REAGAN | All of this chaos makes perfect sense | Wed Apr 09 1997 10:34 | 14 |
| Which one is the default? Always? Per_IO?
If I say $ SET MAGTAPE /SKIP=FILES:2 /FAST_SKIP=ALWAYS, will
that work get me fast skipping for that command?
Why would you want to not use fast skipping when the device supports
it? What problem are you trying to solve to letting you disable
this fast skipping?
As for "per_io", I feel that there is something better we can come
up with. I'll look at a few other DCL commands and see if I can spot
something similar we can plagarize.
-John
|
| 769.2 | more on /fast_skip | STAR::YURYAN | | Wed Apr 09 1997 12:04 | 53 |
| John,
>Which one is the default? Always? Per_IO?
Per_IO is the default
>If I say $ SET MAGTAPE /SKIP=FILES:2 /FAST_SKIP=ALWAYS, will
>that work get me fast skipping for that command?
/SKIP and /FAST_SKIP are two different operations and would not
be allowed on the same command line.
/SKIP indicates *where* to position.
/FAST_SKIP indicates *how* to position.
>Why would you want to not use fast skipping when the device supports
>it? What problem are you trying to solve to letting you disable
>this fast skipping?
* Overhead incurred by using the new method could be significant if
writing a number of backup savesets (for instance) near the end of the
tape. There is positioning backward twice over 2 filemarks, then
forward once, to determine whether or not you're really at a double
filemark, or a filemark and a record.
With the old method, you backup up once over one filemark, and
continue to append from there. There is no further checking/
positioning required.
* Also, in order for this fast search to work correctly, two things
are required. The device must support the SCSI READ_POSITION command
and, it must report blank check at end-of-data. There's no way of
knowing beforehand if the device will report end-of-data even if it
does support the READ_POSITION command.
* Some third party utilities may depend on the documented behavior
of stopping a skip by files on double EOF marks on the tape. To
accomodate this, we use the default of the old skip by records method.
In fact, implementing this support was initiated by 3rd party devices/
applications.
>As for "per_io", I feel that there is something better we can come
>up with. I'll look at a few other DCL commands and see if I can spot
>something similar we can plagarize.
I tend to agree that "per_io" may not be the best choice - so we're
open to suggestions.
Thanks,
M
|
| 769.3 | | TLE::REAGAN | All of this chaos makes perfect sense | Thu Apr 10 1997 13:04 | 12 |
| More questions...
Is /FAST_SKIP permanent? Does it require priv's to execute?
Your explanation of what it is was very clear to me. However, can
you tell me why I would want to say ALWAYS instead of letting
the application do it on a "per_io" basis when it knows what it
wants? Why would I want to "force" this behaviour on an unsuspecting
application?
-John
|
| 769.4 | more answers.. | STAR::YURYAN | | Fri Apr 11 1997 11:31 | 24 |
| > Is /FAST_SKIP permanent?
If you mean, Is this intended to be a temporary workaround ?
This is intended as a permanent feature.
> Does it require priv's to execute?
To utilize fast search will require no more priveleges than any
other SET MAGTAPE command.
>Your explanation of what it is was very clear to me. However, can
>you tell me why I would want to say ALWAYS instead of letting
>the application do it on a "per_io" basis when it knows what it
>wants? Why would I want to "force" this behaviour on an unsuspecting
>application?
Assumedly, there are different applications out there, and levels of
users. Applications may, or may not be modifiable by users. The ability
to set it at will, gives the knowledgable user/application options. It
might also be noted that there are already provisions in BACKUP, COPY
and MTAACP that automatically attempt to use the fast search. This makes
SET MAGTAPE the only user setable option.
|
| 769.5 | | TLE::REAGAN | All of this chaos makes perfect sense | Fri Apr 11 1997 16:18 | 13 |
| No, I meant permanent like you have SET TERM and SET TERM/PERM.
Think about SET MAG /FILE. That only works when you have a real
tape in the drive. Its a one time only thing.
For SET MAG /FAST=ALWAYS, will be be in affect on the next tape, or
the next week or whenver? Or does it revert to the default of
"per_io" when you dismount the unit?
The word "ALWAYS" made me think that the setting would outlast the
process that has the drive mounted.
-John
|
| 769.6 | Is permanent | STAR::YURYAN | | Mon Apr 14 1997 12:19 | 6 |
| Sorry, I mis-understood.
Yes, when these characteristics are set, they stay set unless they
are specifically re-set, or there is a reboot.
M
|
| 769.7 | | TLE::REAGAN | All of this chaos makes perfect sense | Wed Apr 16 1997 13:48 | 7 |
| So, if its permanent, do you need privileges to do it?
I've never seen a SET command that was permanent not require
privileges.
-John
|
| 769.8 | Similar to compaction | STAR::YURYAN | | Thu Apr 17 1997 13:22 | 7 |
| It is permanent, and does not require special privileges. The set
compaction support does the same thing - you don't need extra privs,
and once compaction is set, it stays that way until it is specifically
re-set or the machine reboots.
M,
|
| 769.9 | | TLE::REAGAN | All of this chaos makes perfect sense | Thu Apr 17 1997 15:46 | 14 |
| OK, if you say so. I pesonally find it uncomfortable that a user can do
something to affect the state of the system for subsequent users
without any privileges, but if that's the behavior you want...
Back to the matter at hand, /FAST_SKIP. So its either "its on at the
system level and you don't need to explicitly ask for it" (ie, the
ALWAYS mode) or "you have to manually ask for it on each operation"
(ie, the PER_IO mode). So what do you think of /FAST_SKIP=(ON|MANUAL)
or /FAST_SKIP=(ON|PROGRAMMED)?
-John
|