[OpenAFS-devel] DRAFT: New sysname standard

[Ted McCabe]

At 12:42 PM -0700 4/25/01, Russ Allbery wrote:
>First off, I thought we were going to go with a configurable list of @sys
>names to try in order rather than making people make a bunch of symlinks?
>Or is that still the plan and this is just an outline for the most
>specific possible values?

Yes to the second question.

Here are some concerns that started the development of the sysname standard:
)Parsability.  Scripts and programs can easily interpret the 
different parts of it.
)Readability.  Us humans don't get crossed eyes and/or brains using 
these all the time.
)Well-defined.  A sysname refers to precisely one element in the set 
of {architecture}x{OS,w/version}x{other_relevant_uniquifiers}.

These are the essentials of the concerns.  They evolve from issues 
regarding the use of @sys values within a single cell with varied 
clients and from issues regarding intercell operability.

My observation is that the primary point of discussion, sometimes 
quite heated, has been around what the relevant/correct identifiers 
are in each of the three sets.  I believe one point of the standard 
is to allow for these identifiers to evolve as our various systems 
evolve.

>Phil Moore <Phil.Moore@morganstanley.com> writes:
>
>>      The architecture field describes the CPU, and is typically associated
>>     with a chip set, such as "sun4u" for Sun's UltraSparc. The architecture
>>            field can be interpreted independent of the other3 fields.
>
>[...]
>
>>  +---------------------------------------------------+
>>  |Vendor|Value|             Description              |
>>  |------+-----+--------------------------------------|
>>  |Intel |ia64 |64 bit Intel Chips (Itanium)          |
>>  |      |-----+--------------------------------------|
>>  |      |ia32 |32 bit Intel Chips (Pentium, x86, etc)|
>>  |------+-----+--------------------------------------|
>>  |Sun   |sun4c|Sparcstation 2, IPC, etc.             |
>>  |      |-----+--------------------------------------|
>>  |      |sun4m|Sparcstation 10, VMS bus machines, etc|
>>  |      |-----+--------------------------------------|
>>  |      |sun4u|UltraSparc                            |
>>  |------+-----+--------------------------------------|
>
>It doesn't make a lot of sense to me for you to be distinguishing between
>sun4m and sun4u (and you're missing sun4d, for whatever it's worth) and
>yet not distinguishing between i386, i486, i686, etc.  The level of
>difference is similar, and software compiled for i686 may not run on i386.

Not being a user of Intel chipsets, (that feels like an odd 
qualification?-), it does appear that this is the sort of distinction 
that we want to avoid failing to make.

    --Ted