[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gentoo-project
Subject:    Re: [gentoo-project] Re: [RFC] Proxy-Maintainer heartbeats (re:sunrise/proxy-maint discussion)
From:       Alex Xu <alex_y_xu () yahoo ! ca>
Date:       2015-04-12 22:31:07
Message-ID: 552AF22B.7030607 () yahoo ! ca
[Download RAW message or body]


On 12/04/15 06:15 PM, NP-Hardass wrote:
> Andrew Savchenko <bircoph <at> gentoo.org> writes:
> 
>> I assume you mean a proxied maintainer (a user maintaining a
>> package via a proxy), not a proxy maintainer (a developer proxying
>> packages for users).
> 
> Yes, that is what I meant.
> 
>>
>> If we'll have a per-maintainer ping (e.g. one ping for each
>> maintainer regardless number and state of maintained packages),
>> then we'll end up in a situation, where maintainer is active, but
>> maintains only a subset of assigned packages.
>>
>> If we'll ping each proxied maintainer on per-package level, this
>> will be too irritating (at least from my perspective I'll burn that
>> with fire).
> 
> Regarding per-maintainer or per-package, I'd endorse per maintainer.  My 
> primary goal in proposing it is to deal with those that aren't dealing with 
> maintenance at all. It's likely that if a user is still actively maintaining 
> one package that they are actively maintaining their others or are more 
> likely to respond to an inquiry (because they are actively involved 
> elsewhere in the project).  Additionally, for a per-maintainer, the ping 
> email could contain all packages for which they are listed, and they are to 
> respond to whether they are actively managing all, some, or none. (obviously 
> lack of response implies none)  If some, then they can be handled 
> appropriately (though I haven't thought that through).
> 
> I'm expecting that this would be automated: scan the tree, grab info via 
> proxy maintainer metadata tag, send ping, take action accordingly.
> (To make that work most efficiently, adding an optional attribute to 
> maintainer, proxy="yes" or "no")
> 
>>
>> I see no better alternative to ping-before-takeover. Though as you
>> should remember from openafs experience, this takes time (which is
>> also irritating). But we almost never know why someone is
>> unavailable, so it will be impolite to kick people if they are
>> out of reach for a month or two.
> 
> Ah, but with OpenAFS, the proxied-maintainer was out of touch for a year, 
> based on bug dates and OpenAFS release dates.  I shudder to think how many 
> other packages out there might be sitting with proxied-maintainers that are 
> no longer involved.  I recently took over another package which was sitting 
> for over a year, and I managed to contact the previous proxied-maintainer 
> who said that he wasn't even using Gentoo anymore.
> 
> Thanks for your input.  Looking for as much as people are willing to give.
> 
> --
> NP-Hardass
> 
> 

I think a major problem here is that it's very easy to forget about
bugs; the Mozilla Bugzilla has needinfo? etc flags and defaults to
sending notifications if they are left alone for too long.

More broadly, that feature makes it possible for someone to track their
own involvement in bugs; since we usually assign bugs to the package
maintainer, someone intending to create a patch can easily forget about
it. It is difficult to search for bugs where this has happened, since
the only field that is usable for this is CC, where each person is most
likely only there to track progress, not necessarily contribute themselves.


["signature.asc" (application/pgp-signature)]

[prev in list] [next in list] [prev in thread] [next in thread]