[2265] in testers
sun4 7.6F: $bindir and @sys on solaris
daemon@ATHENA.MIT.EDU (Derek Atkins)
Wed Jul 28 21:12:00 1993
To: sparc@MIT.EDU, testers@MIT.EDU
Cc: bug-sparc@MIT.EDU, bug-afs@MIT.EDU, warlord@MIT.EDU
Date: Wed, 28 Jul 93 21:11:47 EDT
From: Derek Atkins <warlord@MIT.EDU>
System name: deathtongue
Type and version: SPARC/IPX 7.6F
Display type: cgthree
What were you trying to do?
Use $bindir or @sys values on sparcstations.
What's wrong:
$bindir has the wrong value under solaris, and @sys is
inconsistent across platforms.
What should have happened:
$bindir under solaris should have a different value than
$bindir under SunOS 4. And the @sys values for classics should
reflect that they are a sun4m, not sun4c platform.
Please describe any relevant documentation references:
I realize that this is a religious war over the naming, but
this is seriously a big problem. There are a few problems that this
causes, and a few reasons why this should be changed. Let me start
with the @sys values:
There are many different @sys values across sun platforms,
which reflect changes in the architecture and operating system. One
reason that this was done is because the kernel modifications for AFS
require different patches depending on the machine type you are using,
and the different OS releases from Sun also require different types of
modifications. Hence, there are many different @sys values for the
sun architecture, but they all exist for a reason, and it can cause
problems to use a 4m kernel mod on a 4c machine, or vice-versa. I.e.,
A classic vs. an IPX. As an example:
UNIX(r) System V Release 4.0 (ISLA-DE-ENCANTA)
Athena Workstation (SUN4m) Version 7.6F Fri Jul 23 21:38:02 EDT 1993
...
> fs sysname
Current sysname is 'sun4c_51'
Athena Workstation (SUN4c) Version 7.6F Tue Jul 27 12:29:06 EDT 1993
(deathtongue-warlord) in /mit/warlord
Ready> fs sysname
Current sysname is 'sun4c_51'
Now, for the second problem, the $bindir inconsistency. When
Athena first got the Suns, it used a $bindir of "sun4" so you could
have a "sun4bin" which contains your sun binaries. This was fine.
But now that Athena has switched to solaris, it still kept the name
"sun4", which is a problem!!!!
The point im trying to make is that its ok to have different
entities have the same name so long as things that depend on that name
work on both entities... for example, the DS3100, Maxine, and
5000/133 are all different DECstation machines, but they are all
binary compatible. Same goes with the Sun machines Sun4, Sun4c, and
Sun4m, so long as they are all running either SunOS4 or SunOS5..
However, When you have the incompatibility (i.e., from SunOS4 -
SunOS5) then you should create a new name.
As an example, the SIPB would like to support both SunOS and
Solaris clients, but we *cannot* support both machines because of the
$bindir inconsistency across incompatible operating systems.
PLEASE PLEASE PLEASE change the $bindir under solaris to
something other than "sun4bin". I recommend "solarisbin" or
"sunsolarisbin" or "sun5bin" something like that. Anything to
differentiate it from the "sun4bin" set by precedent for SunOS 4.
I look forward to seeing this change in 7.6G!
Thanks!
-derek
