Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 20 Apr 95 16:02:29 MDT
From:      terry@cs.weber.edu (Terry Lambert)
To:        gene@starkhome.cs.sunysb.edu (Gene Stark)
Cc:        hackers@FreeBSD.org
Subject:   Re: Minutes of the Thursday, April 13th core team meeting in Berkeley.
Message-ID:  <9504202202.AA28852@cs.weber.edu>
In-Reply-To: <199504202110.RAA08175@starkhome.cs.sunysb.edu> from "Gene Stark" at Apr 20, 95 05:10:31 pm

next in thread | previous in thread | raw e-mail | index | archive | help
> >It was not the result of problems endemic to a release process.
> 
> I disagree with this, as the prior release that was burned into CD-ROM
> (1.1) IMHO was also an inferior release due to a rushed testing phase.
> Release 1.1.5.1, which is the best release to date, was essentially the
> result of testing and bug fixes applied to 1.1.

Let me clarify the term "endemic".  I do not believe that it is an
inevitable outcome of the release process as it is defined.

In other words, it is a result of failure of process rather than the
process itself.

It's not a question of procedure, it's a question of failure to follow
procedure.

> Every time the beta phase of a release has been shrunk to zero, I have
> complained.  I usually see everyone say "I'll *never* allow this to happen
> again," but by the time the of the next release, the same thing has
> happened again.  It is my impression that the same thing is happening
> right now with 2.0.5.

A shrinking beta phase is a violation of the release protocol as it has
been agreed to.  The unpleasent part of release engineering is in
being the source tree hall monitor -- enforcing protocol.

> My feeling is that a system version from approximately April 7 would be
> a good one to use as a base for 2.0.5.  Check out that release, put it
> out for beta test, integrate bug fixes only, and burn it into the CD-ROM.

This is a good rollback date (at least as good as any other) to pick as
the code freeze date after which new features will not be introduced to
the release branch.

I'm sure, though, that Paul (or whoever has replaced him as release
engineer) has already branched the code at what looked like a logical
breakpoint.  As long as a branching has taken place, it's really not
kosher to second guess him unless we are willing to walk a mile in
his shoes.

I would like to see release packaging polished to the point that a beta
release can be rolled from a full install machine by typing "make release"
somewhere so that we can get an immediate feel for beta (or so that
those of us with extra machines can incrementally test and source tree
that fully builds).

This is an unreasonable requirement to have at this time without the
support already in place, but it is certainly a topic that should be
taken up in the near future.  Unfortunately, a lot of what takes place
in an install is still hand massaged, which means that there is a
terrific investment in the people doing the work instead of in tools
to do the work automatically -- this leaves the process particularly
susceptable to disruption by distraction of the people or the people
being hit by a bus, or snipers from The Forces of DOS taking them out.

I think that many of the complaints that you have are valid relative
to 2.0, but that they are indicative of failure to follow procedure
more than anything else.  Hopefully, they will be proven invalid for
the next and subsequent releases, which are not under the shadow of
exceptional circumstances (USL, etc) that 2.0 labored under.


					Terry Lambert
					terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9504202202.AA28852>