Skip to Content.
Sympa Menu

sm-discuss - Re: [SM-Discuss] stable/testing branches

sm-discuss AT lists.ibiblio.org

Subject: Public SourceMage Discussion List

List archive

Chronological Thread  
  • From: Ari Steinberg <ari.steinberg AT stanford.edu>
  • To: Eric Womack <eric AT lasvegasdata.com>
  • Cc: sm-discuss AT lists.ibiblio.org
  • Subject: Re: [SM-Discuss] stable/testing branches
  • Date: Fri, 26 Jul 2002 18:19:48 -0400

my last comment (probably) on this thread...

At 11:57 AM 7/25/2002 -0700, Eric Womack wrote:
But even to say that we will determine if a package is is stable because
Debian says it's stable is too much. WE need to test is and determine for
ourselves if something is stable.

I'm still not convinced that this is actually what's happening. How much testing is really involved before something becomes a part of the test/stable branch? It seems more to me like once a version has been sitting around for a week or two without any complaints it's automatically merged into the test/stable branch. How do we know, however, that it truly has been tested, and that nobody actually did have a problem with it? The current process, if I'm not mistaken, is simply for one person (Tony) to deal with all of this for two branches, but I think it is too much to expect of him to say that he has *pro-actively* verified that two grimoires work perfectly. Instead, I think we simply designate things as stable in a reactionary way - if nobody complains then we assume, and assuming of course is not a good thing.

I guess if the idea of relying on Debian seems bad, then at the very least it would seem to make a bit more sense to me for section maintainers to track 3 versions of each program (stable/testing/devel), so that the people deciding what is stable can have a little bit more focus in making this very difficult decision.

The whole reason I started thinking about this was after reading a post somewhere on Slashdot about how Debian actually tracks bugs not just in their own tools but in all of the programs that they have packages for, and has admirably high standards for what is truly considered "stable". I think this approach makes a lot of sense, while ours seems a bit more misleading (if I am a sysadmin who uses a "stable" grimoire, it probably means that stability is crucial to me, and I'm definitely not convinced that this sysadmin is getting something that can truly be considered stable).

I've already suggested getting the help of Debian, and have also suggested instead adding the extra job of tracking 3 version #'s to a section maintainer's responsibilities. My final idea is that perhaps stable should truly be "stable" in terms of not changing...maybe it would make sense for stable to be a grimoire that is intended to not be updated - just download the stable iso, which has verified version numbers, and don't plan to update until the next stable iso is released (with the possible exception of important security fixes), with releases happening maybe once every couple months. These stable iso's could even come in a multi-disc version with all the source for every program included on the discs (hmm - not sure how many discs that would end up being).

Just some things to think about.
-Ari





Archive powered by MHonArc 2.6.24.

Top of Page