Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 9 Oct 1997 13:53:48 -0500
From:      Richard Wackerbarth <rkw@dataplex.net>
To:        "Rodney W. Grimes" <rgrimes@GndRsh.aac.dev.com>
Cc:        stable@FreeBSD.ORG
Subject:   Re: Fwd: CVSup release identity
Message-ID:  <l03110703b062cdb61a4f@[204.69.236.50]>
In-Reply-To: <199710091732.KAA03880@GndRsh.aac.dev.com>
References:  <l03110700b05eddbc3dc5@[204.69.236.50]> from Richard Wackerbarth at "Oct 6, 97 01:39:02 pm"

next in thread | previous in thread | raw e-mail | index | archive | help
>> >On Mon, 6 Oct 1997, Richard Wackerbarth wrote:
>You're only looking at 2 of many values, your making your decision based
>upon incomplete information.  Values of ``BRANCH'' (to be renamed FLAVOR
>after the 2.2.5 release) are ALPHA, BETA, ..., RELEASE, STABLE, CURRENT.

I view it in a slightly different manner. I agree that ``BRANCH'' is a poor
term. I think that we agree that the BRANCHes are really 2.1, 2.2, 3.0, etc.
I also agree that there are values other than STABLE and CURRENT.

However, these values represent a PHASE of development. For example,

Development (pre-release)
Alpha (testing)
Beta
Released
Patched

might be terms that describe the situation. (I'm not arguing the names,
just look at the concept). The important thing is that any given branch can
have only one phase at any given time. Further, there is fundamentally
nothing sacred about the transition from one phase to another. If you have
continuous access to the development tree, the quality of the code
immediately before the transition is the same as it is immediately after
the transition. The only thing that is important about the phase is that
there are some implied expectations and protocol restrictions on the
developers which are organizational policy.

In addition, there are certain snapshots of the development progression.
These are called releases. Someone packages them up and distributes them to
users. These users cannot change them (without creating something else
which is no longer the same and therefore should not have the same name).

Anyone who assembles a system between these named snapshots simply has a
version of the branch based on its contents at the time the snapshot was
extracted.

The development phase is not important. Only the point on the development
timeline matters. (Did you build your system from sources that were
extracted before or after some change that was introduced at xxxxx?)

>Now, tell me it can go away since you now have all the information and
>I't will after the 2.2.5 release.
>
>> The "current" branch is presently called "3.0".
>
>Okay, I'll buy that, it means we can set FLAVOR == CURRENT instead of
>FLAVOR == STABLE on the RELENG_2_1 and RELENG_2_2 branches.  Now is that
>ever going to confuse the shit out of users :-)

Yep! Anyone who builds a system from the head of a branch is building the
"current" version of THAT branch (current as of a particular time). In this
context, "CURRENT" is meaningless. Rather, we should be DELETING both of
these designations.

The only value in identifying a non-release system with anything other than
the timestamp is that it may be easier to remember that a change occured
between a pair of named releases rather than remembering that exactly when
it occurred. However, this only helps to exclude systems that fall outside
the named bounds. Between the bounds, you still have to look to the more
specific timestamp.

My proposed nomenclature is to designate the branch, the last release tag
along that branch, and the timestamp of the extraction.

Communication about a particular branch would be identified by the branch
and not by the flavor. I can see it now.  As we are in aa alpha period, all
discussion should be on the "alpha" mailing list. Next week, please move
the discussion to the "beta" list...  I don't think so.  If we were a much
larger organization and the "users" had access to ONLY the released
versions of development, you might make an argument for it. However, we are
too small and everyone has access to "today's" version. One list per branch
should be sufficient.

Richard Wackerbarth





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?l03110703b062cdb61a4f>