| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1674.1 | | COOKIE::LAWSON | Carpe Heli Diem, eh? | Sat Sep 14 1996 13:16 | 14 |
| 1674.2 | $SETENV not implemented on these systems! | PADBOL::STEFF | Marc Stefanelli, #858-5282 | Sat Sep 14 1996 15:51 | 17 |
| 1674.3 | | PADSOU::STEFF | Marc Stefanelli, #858-5282 | Tue Feb 11 1997 06:57 | 22 |
| John,
I IPMTed this one against AlphaStations: it took ages to make any
progress on it.
Finally, the proper way to set the BOOTCMD_DEV variable is to use
SCSI 0... string (as for BOOTED_DEV) instead of DKA0.
Engineering claims that this implementation complies to SRM specs
and that is the job of code using the console interface to present
the right string! I dunno who is right regarding the specs and even
I would prefer to have a consistent interface between console prompt
and console routines.
The real issue is that there is at least two implementations out
there and no idea on what to implement something being able to handle
both cases.
- Marc S.
|
| 1674.4 | | COOKIE::LAWSON | Carpe Heli Diem, eh? | Fri Feb 14 1997 18:15 | 12 |
| Welcome to the world of the "standardized console interface". The idea, when the EVAX (now Alpha)
was proposed is that we'd avoid the absolute nightmare of different interfaces that the VAX had
to support, by standardizing the interface. It did get standardized. There is a consistent way
to pass strings from the operating-system to the console.
The problem is that noone bothered to standardize the strings that get passed.
The worst problem is that in many cases, there is no way for the operating-system to "figure out"
the correct string needed by the console. The console can figure out what device it means from
its string, but only the console can figure that out.
I'm afraid we're not going to get anywhere on this one ... I'm sorry.
|
| 1674.5 | a proposal for your next ECO | PADSOU::STEFF | Marc Stefanelli, #858-5282 | Thu Feb 20 1997 04:52 | 29 |
| When operating properly an ISL does not really need BOOTDEF_DEV to
be set.
An interim solution would be to implement a logical name defined on
the server side in order to avoid setting this variable.
Better solution require to get a formal statements from console
designer on what string flavor should be provided and how to get
the `flavor' style of a specific platform.
Actually we have at least to kind of console lets says TurboChannel
(TC) and PCI ones behaving differently.
From various testing:
- DEC3000-300 (TC) seems to accept both string formats : I will have
access to a DEC3000-300 next week so I can confirm this. But I
dunno where to get a formal statement on this.
- DEC4000/AS2100/AS200 seems to behave the same: clears NV variable
on improper string value.
If both implementation accepts the same string format, RSM can easily
manage to set BOOTDEF_DEV same as BOOT_DEV: both variables are set in
the same operations. The real question is 'can we rely on this on the
long term or should we prepare to see other flavors coming out?'
- Marc S.
|