Proposed: configuration management for 3b1/7300/unixpc

Kris A. Kugel kak at hico2.UUCP
Sat Apr 6 08:47:44 AEST 1991


In some ways, the 3B1/unixpc is quickly becoming a hobbist's machine.
In part, because the upgrades are hobbist's upgrades, but also
because these machines are diverging more and more from the distributed
standard.

I can't figure out what's wrong with anybody else's machine, because
each machine is becoming unique in its software and hardware.

The recent discussion about porting a new os, the discussion on ttys
by Karl Swartz and Thad Floryan, and some recent querys posted seem
to indicate a need to be able to specify what is on a particular machine -
what hardware, what software.

There are several hardware patches floating about, including HD2,
720meg floppy, PS5..1, vidpal.  There are several versions of the
OS software, 3.0, 3.50, 3.51, 3.51m.  There are several other
system-modifying packages, kernel serial patch, DST patch, HDB uucp,
math package upgrade.  Others are possible: ua removal, security upgrade. . .

I think that we need to start considering how to handle configuration
management for this system, before nobody can understand anybody else
in this group.  I think we need a more formal way of describing
"how, starting with a vanilla 3b1 and version X of the OS,
 we get Fred's machine"

Otherwise, any possible future releases of the OS, or amateur releases
of software of other sorts will have difficulty avoiding interfering
with other changes.

I still think we should come up with a set of model systems,
each a description of a set of hardware and/or software.
Like a "list of ingredients" in a recipe.
And maybe a new way of designating systems: 3.51m-HD2/720/VIDP-sp/mp/hdb.
And proper dependencys:
"this change requires. . ."
"this change removes. . ."
"this change replaces . . ."

In some way, the installable package idea of the ua seems like a good
start.  But that approach seems too limited, with too many small pieces
to deal with, unless we start bundling together our own packages:
	"Standard Kernel patches 3.51m-A.0"
	"Standard Library modes 2.1"
	"Standard g++ installation"
	"Standard C-news installation"

I dunno, but I used to have a reasonably clear picture of what somebody
had when they said they had a 3b1.  Now, that picture is getting fuzzier
and fuzzier.

I know that there are people out there who can design and build a computer
with breadboards, or/and write an OS with device drivers from scratch,
but I think that the majority of the users of 3b1's are not such people,
and even among those who can, who has the time?

Maybe clarifing the 3b1 world is not possible in an environment
such as ours, where there is no longer any central authority
organizing releases.  And organizing this stuff will require effort.

But the alternative seems to me to be increasing confusion,
as machines diverge farther and farther apart, and we have
no way of easily qualifying the changes.

Maybe the 3b1 is not long for the world, and this will meerly delay
its fate.  But I'd like to get a couple of years more use out of my
machines, and would like to think that they will be useful to somebody
else after I'm done with them.

So I think this is worth discussing.

                               Kris A. Kugel
                             ( 908 ) 842-2707
                    {daver,ditka,galaxia,zinn,zorch}!hico2!kak
                      internet: kak at hico2.westmark.com
                         uunet!westmark!hico2!kak
                      uunet!tsdiag.ccur.com!hico2!kak (maybe)



More information about the Comp.sys.3b1 mailing list