Jason, I understand what you're saying, and from a process point of
view, I somewhat agree. From a tool point of view though, that's not
true. Most SCM systems, and I can speak specifically for (in recent
usage anyway) CVS, Perforce, and Subversion, the tools support deleting
tags or in some manner reassociating the files/versions associated with
the tags. Because of this, it's someone dangerous to consider them as
immutable from a process point of view, because underlying things at the
tool level someone could back door you.
It's almost better to build something into your process to recognize
what the tool is capable of, so you don't run into trouble. As a
possibility (not saying it currently works or should work like this) it
might be better as what Andrew says, tag a revision number in subversion
and then vote on that revision number really being the GA release. If
the vote is turned down, then untag it, and tag a future voted revision
number. Now, you're in concert with what the tool supports. Even if
someone goes behind the scenes and changes the tag, if you have a 'vote
history' saying that a certain revision number should equal a certain
tag, then you can always 'fix' the tag later. The way this kind of
approach had to be done in CVS was simply using 'timestamps'.
Anyway, I guess this is a long way of saying I agree with Mr. Petro.
---- Cris J H
Andrew Petro wrote:
Jason,
* In the future, we should likely vote against a specific changeset or
milestone being tagged as a GA, not on releasing the GA tag. Afterall,
if we voted no what would we do? Retag GA but pointed at a different
revision number? Seems to violate the contract of a tag if it changes
over the lifetime of the repository.
Yes, if we'd voted no I'd have removed the tag and re-tagged with
whatever changes were required to make the release acceptable. It's a
harmless violation.
In theory I could have tagged and released an RC3 and proposed that the
minimum diff on that be accepted as GA, but that sounds like extra work
for benefits that exist only in principle.
Andrew
--
You are currently subscribed to [email protected] as: [EMAIL
PROTECTED]
To unsubscribe, change settings or access archives, see
http://www.ja-sig.org/wiki/display/JSG/uportal-dev
