On Sat, Feb 21, 2009 at 5:55 PM, Assaf Arkin <[email protected]> wrote:
> On Fri, Feb 20, 2009 at 7:27 PM, Matthieu Riou <[email protected]>wrote: > >> On Fri, Feb 20, 2009 at 3:48 PM, Assaf Arkin <[email protected]> wrote: >> >>> On Fri, Feb 20, 2009 at 2:24 PM, Matthieu Riou >>> <[email protected]>wrote: >>> >>>> On Fri, Feb 20, 2009 at 2:08 PM, Assaf Arkin <[email protected]> wrote: >>>> >>>>> >>>>> In my scenario there's an RC, a separate entity from the final release. >>>>> And it may have passed through the release process and voted on. Or not. I >>>>> didn't specify because it doesn't seem to matter. >>>>> >>>>> So "not kosher" seems to me like inventing one very specific process >>>>> and using it as strawman to argue that RC is problematic by nature. >>>>> >>>> >>>> Of course, I'm not disputing that. The problem is with a very specific >>>> way to handle RCs (vote on it and then just release from the same >>>> tag/branch >>>> without re-voting) not with RCs in general. >>>> >>> >>> So I don't understand what you mean when you say "RCs aren't fully >>> kosher". If there's nothing inherently non-kosher about RCs, and we're not >>> specifically discussing a particular non-kosher process, why not consider >>> doing RCs? >>> >> >> I was referring back to the thread I linked to earlier ([1]) and the >> related discussions that happened in legal discuss (also linked in that >> thread). So when I was talking about RCs it was in that context (again, >> voted RCs released as final from the same tag/branch without re-voting). >> >> I don't have anything against RCs as an industry practice. We (as an >> Apache project) just need to handle them carefully because of the release >> process. With that process in mind, the possibilities I see are: >> >> - Vote on RCs and final releases. Longer, heavier process (2x72h). >> - Only vote for final releases. The RCs are therefore unofficial and >> not meant for wide consumption. >> - We could also choose to not have releases tagged as RCs and skip >> numbers as a result. If a release is good, we vote on it, otherwise we cut >> the next. >> >> What I like about the last option for ODE is that it's simple. But we can >> probably make the second one work too if jumping numbers is A Big Deal. >> > > It's simple until something goes wrong and then people start complaining > about the quality of the releases. The thing is, people don't appreciate > "release often" as much as they do "release well tested code". If you keep > jumping versions when things break, people start treating you as a source of > broken releases. So RC becomes this period of giving the code one last > chance to prove it's well tested. > > How involved the RC process should be depends on the likelihood and > consequences of things going group, coupled with the time it takes to find > them: > > - Little consequence, easy to find. The code you're voting to release is > the only candidate you need, and 72 hours enough to judge its quality. > - Serious consequence, easy to find. 72 hours not enough, you need an RC > that is distinct from the final release, and available for longer period. > - Serious consequences, hard to find. Get more people involved by releasing > an RC (vote and all) and giving people time to test it out (could take > weeks). > > The last process is the heaviest, but it also guarantees more people > looking at the code before cutting the final release. How likely I am to > look at an RC is correlated to how long the RC window is, and how easy it is > to obtain the RC, and I wouldn't be surprised if that's the general rule for > other people. > > Judging these processes in isolation is meaningless, the question is, which > one is right for Ode? > We have 2 branches with different levels of maturity. For 1.X, which is stable (sort of), I was gearing toward the first option, which is why I went directly for a 1.3. For the trunk, we're clearly on the second option (releasing betas at this point). Matthieu > > Assaf > > > >> >> >> Matthieu >> >> [1] http://markmail.org/thread/o73bu7mo2tqnrv2p >> >> >>> Assaf >>> >>> >>>> >>>> Matthieu >>>> >>>> >>>>> >>>>> Assaf >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> Matthie >>>>> >>>>> >>>>>> >>>>>> Assaf >>>>>> >>>>>> [1] >>>>>> <http://en.wikipedia.org/wiki/Software_release_life_cycle#Release_candidate> >>>>>> http://en.wikipedia.org/wiki/Software_release_life_cycle#Release_candidate >>>>>> >>>>>> >>>>>>> >>>>>>> The former involves further delays, a heavier process, ... >>>>>>> >>>>>>> Matthieu >>>>>>> >>>>>>> >>>>>>> > >>>>>>> > Thanks, >>>>>>> > Milinda >>>>>>> > >>>>>>> > On Thu, Feb 19, 2009 at 10:31 PM, Ciaran < <[email protected]> >>>>>>> [email protected]> wrote: >>>>>>> > >>>>>>> > > On Thu, Feb 19, 2009 at 4:54 PM, Assaf Arkin <<[email protected]> >>>>>>> [email protected]> wrote: >>>>>>> > > >>>>>>> > > > On Thu, Feb 19, 2009 at 8:43 AM, Alex Boisvert >>>>>>> > > > <<[email protected]> >>>>>>> [email protected]> >>>>>>> > > > wrote: >>>>>>> > > > >>>>>>> > > > > The confusion comes from the fact that we pseudo-released >>>>>>> 1.3. It >>>>>>> > > should >>>>>>> > > > > have been a RC1. >>>>>>> > > > > >>>>>>> > > > > I don't think it's a good idea to use version number without >>>>>>> > qualifiers >>>>>>> > > > if >>>>>>> > > > > they are not real releases. Now version 1.3 has been >>>>>>> "released" but >>>>>>> > > > > there's >>>>>>> > > > > no mention of it on the web site, there was no vote, etc. >>>>>>> > > > >>>>>>> > > > >>>>>>> > > > The question is: does anyone have a copy they're using, >>>>>>> thinking it's >>>>>>> > the >>>>>>> > > > official 1.3 release? >>>>>>> > > > >>>>>>> > > > >>>>>>> > > > > >>>>>>> > > > > >>>>>>> > > > > The first question many people will have when they download >>>>>>> 1.3.1 is >>>>>>> > > > "What >>>>>>> > > > > happened to 1.3?" >>>>>>> > > > >>>>>>> > > > >>>>>>> > > > 1.3.1 >>>>>>> > > > * Fixed issue with packaging, new version no. to remove >>>>>>> confusing with >>>>>>> > > > pulled-back release 1.3. >>>>>>> > > > >>>>>>> > > > 1.3 >>>>>>> > > > * Pulled back due to issue with packaging. >>>>>>> > > > >>>>>>> > > > Assaf >>>>>>> > > > >>>>>>> > > > >>>>>>> > > > > >>>>>>> > > > > >>>>>>> > > > > alex >>>>>>> > > > > >>>>>>> > > > > >>>>>>> > > > > On Thu, Feb 19, 2009 at 8:12 AM, Matthieu Riou < >>>>>>> > > <[email protected]>[email protected] >>>>>>> > > > > >wrote: >>>>>>> > > > > >>>>>>> > > > > > On Wed, Feb 18, 2009 at 6:24 PM, Alex Boisvert < >>>>>>> > <[email protected]>[email protected] >>>>>>> > > > > >wrote: >>>>>>> > > > > > >>>>>>> > > > > >> On Wed, Feb 18, 2009 at 4:55 PM, Matthieu Riou < >>>>>>> > > > <[email protected]>[email protected] >>>>>>> > > > > >wrote: >>>>>>> > > > > >> >>>>>>> > > > > >>> PS: Did you mean "Cut a new 1.3 release" ? >>>>>>> > > > > >>>> >>>>>>> > > > > >>> >>>>>>> > > > > >>> Mmh no, I've already cut 1.3 and if we re-release it's >>>>>>> going to >>>>>>> > be >>>>>>> > > a >>>>>>> > > > > new >>>>>>> > > > > >>> version number, otherwise we'll end up with some >>>>>>> confusion. Hence >>>>>>> > > > > 1.3.1. >>>>>>> > > > > >>> >>>>>>> > > > > >> >>>>>>> > > > > >> I guess I'm already confused... :-| 1.3 was not >>>>>>> officially >>>>>>> > > released >>>>>>> > > > so >>>>>>> > > > > >> where's the harm? >>>>>>> > > > > >> >>>>>>> > > > > > >>>>>>> > > > > > A few people already downloaded it and tried it. That's a >>>>>>> first >>>>>>> > > chance >>>>>>> > > > of >>>>>>> > > > > > confusion. And later when we'll ask "which version are you >>>>>>> > running?" >>>>>>> > > > and >>>>>>> > > > > the >>>>>>> > > > > > answer is 1.3, which 1.3 does that mean? Version numbers >>>>>>> are cheap. >>>>>>> > > > > > >>>>>>> > > > > > I remember we had a similar discussion some time ago on >>>>>>> this ML >>>>>>> > about >>>>>>> > > > 1.2 >>>>>>> > > > > > or 1.1, we re-released the same version but the consensus >>>>>>> back then >>>>>>> > > was >>>>>>> > > > > that >>>>>>> > > > > > it was "wrong" :) >>>>>>> > > > > > >>>>>>> > > > > > Matthieu >>>>>>> > > > > > >>>>>>> > > > > > >>>>>>> > > > > >> >>>>>>> > > > > >> alex >>>>>>> > > > > >> >>>>>>> > > > > > >>>>>>> > > > > > >>>>>>> > > > > >>>>>>> > > > >>>>>>> > > I don't mind as long as *something* is released <g>... although >>>>>>> today our >>>>>>> > > testing has flagged up some issues around XSL in BPEL that used >>>>>>> to work, >>>>>>> > > still trying to diagnose :( >>>>>>> > > - Cj. >>>>>>> > > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > -- >>>>>>> > <http://mpathirage.com>http://mpathirage.com >>>>>>> > <http://wso2.org>http://wso2.org "Oxygen for Web Service >>>>>>> Developers" >>>>>>> > <http://wsaxc.blogspot.com>http://wsaxc.blogspot.com "Web Services >>>>>>> With Axis2/C" >>>>>>> > >>>>>>> >>>>>> >>>>>> >>>>> >>>> >>> >> >
