Re: Summary of CUT discussions
Joey Hess, 2010-09-27 15:26:10 -0400 :
> Roland Mas wrote:
>> At least for some packages, it's hard enough ensuring a more-or-less
>> pleasant experience in a stable release; trying to provide it on a
>> moving target is *much* more work, especially if one must support
>> upgrades from any version younger than X months (as has been
>> suggested). This is not something that I can currently support, and
>> it shouldn't be something that “we” can claim to support because that
>> would be a lie. And if, as a maintainer, I'm told by whatever
>> powers-that-be that I need to care for this use case from now on, I'm
>> not likely to take that with a smile.
> Well, we know that fully 27% of popcon-reporting users already use
> unstable or testing. So in general, developers already have an incentive
> to keep unstable and testing usable for those users, not just stable.
I'm fine with an incentive. An official promise by the project that
unstable and testing (or rolling) *will* be usable, on the other hand,
makes me really nervous.
> There are always going to be special cases, like gforge/fusionforge,
> which I assume is what you're specifically talking about.
Good guess :-)
> If I'm reading this graph right, there are roughly no users of
> fusionforge yet. Obviously it doesn't make sense for you to spend time
> on unstable-to-unstable upgrades.
> I wouldn't be at all surprised if CUT ended up being mostly used for
> desktop systems, and not for servers, since the desktop is exactly the
> area where rolling releases with constant shiny stuff and new hardware
> support is most needed.
Certainly. And CUT is very probably going to be useful for them, just
as testing can also be useful. However, from what I understand,
CUT/rolling is going to be basically testing plus chromium
(oversimplified). I'm completely fine with that, as it's just another
suite that flows from unstable. The problem arises from what we, as a
project, tell the world CUT is. If we want to call it “safe for many
users” or “supported”, which is as far as I know its most important
selling point, then we need to take out some packages. I don't know how
many, but at least some; fusionforge is admittedly pathological, but I
wouldn't be surprised about other applications that use a database and
need to care about data migration. Wikis, webapps, ERPs, collection
management stuff, that sort of thing. Do we need those in CUT? I don't
know. But CUT suddenly becomes testing plus chromium minus a number of
packages. If that still fits the goals of CUT, then by all means go
ahead; I was just reacting to the stance that it's only a communication
issue, because it clearly isn't.
Neko-no me-to, onna-gokoro-to, aki-no-sora. -- Proverbe japonais
(« Souvent femme varie, bien fol est qui s'y fie. »)