On 4/5/07, Allen Gilliland <[EMAIL PROTECTED]> wrote:

Dave wrote:
> In the past our release criteria was "get three Incubator PMC members
> to vote +1"
>
> What is the standard now? Here are some questions to consider.

good question.

the standard is three binding votes from the committee responsible to
the board for the code. that allows the release manager to act on
behalf of apache rather than as an individual. in the event of
litigation apache, would then be able to defend the release manager.

when roller was in the incubator, the responsible PMC was the IPMC.
now that roller is a top level project, the roller PMC is responsible
for the code and so 3 binding votes are required by roller PMC
members.

<snip>

> Are you allowed to vote +1 if you have not downloaded, installed and
> at least done a quick sanity test (e.g. posting entries, trying a new
> feature or two) on the release files? I'm pretty sure the answer is
> no.

I agree, voting should be reserved for people who actually tested, not
just anyone who is a committer and is say "okay" to the release.

in the end, it's the personal reputation of each committer on the line
as well as the collective reputation of the roller PMC. releases with
difficulties may be held up in public as anti-patterns for years to
come.

the board charges the roller PMC with oversight of the codebase. this
is particularly important when it comes to releases. those whose votes
are not binding can apply whatever criteria they wish but roller PMC
members should check the release carefully before voting.

> When a new RC is released does that invalidate existing votes on the
> release? Not sure about this one. I guess it depends on the extent of
> the changes. When I'm doing a release I want at least one person other
> than me to sanity test the build.

tough call, but i would say no.  if an RC has 5 bug fixes and 2 new
features and all but one of them is tested, but that last one requires a
code fix and updated RC then that shouldn't invalidate all previous
votes.  However, each time a new RC is created there should at least be
a few people to d/l that specific RC and run through a basic deployment.

IMHO it's cleanest to hold a new vote for each candidate

- robert

Reply via email to