New port with maintainer ports@FreeBSD.org [was: Question about maintainers]

[Mark Linimon]

On Fri, Jul 29, 2005 at 12:56:50AM +0200, Roman Neuhauser wrote:
> So, where do we end? With MAINTAINER=ports at FreeBSD.org ports.
> Where's the difference?

The difference is that we don't just give up and start out that way.
There's at least _someone_ who will stand behind the code that got
committed to the tree, for at least _some_ period in time.

> "looking after it"? There's someone who hasn't reacted to PRs for
> at least several months, someone "the users" (emphasis added) can
> ask questions of in vain, even if the answer always is (two months
> later) "maintainer timeout".

Yes, there are maintainers who need to be reset.  I have already made
that point.

> > Most users don't expect "mostly working".  They look at this list,
> > see that they can install something from it, and then get aggravated
> > when they find out they've wasted their time trying to use it.
> 
> I have written several different responses to this paragraph,
> and deleted them all. Let me ask a question instead: can you
> show me some hard facts about those mysterious "most users"?

No, just the last 2-3 years of reading through postings to the various
ports-related mailing lists.  I do not save all the postings that I've
replied to trying to help people.

> You know, I'd really like to see a proof that most of the FreeBSD
> users are what you call them: *users* (emphasis added to point out
> the derogative).

That's your derogative -- not mine.  Please don't misattribute quotes
to me or put words in my mouth, thanks.  This thread is problematic
enough without that.

I mean "user" in this case to be someone who, without necessarily having
an understanding the mechanisms by which the Ports Collection works, sees
something they want to install because it's neat, and tries to do so.

> And if you do manage to prove that most users
> of FreeBSD are what you imply: people who can't read a simple
> Makefile to find out the dependencies for example, people who
> can't fix a simple problem in a port Makefile they wanted to use,
> then I'll concede it's time to look elsewhere.

Your model of FreeBSD seems to be that everyone has at least a medium
technical level of expertise.  But I see people trying to use FreeBSD
where that's not the case.  There is plenty of evidence in GNATS.
However, even if this turns into a What Is FreeBSD thread ("only for
experts?  something for everyone?"), I won't continue to follow up to
it, because all that's done is to fill up various mailing lists before,
and it's something that is never going to come to consensus.

No, I don't think that _most_ users of FreeBSD have less than a medium
level of techical expertise, so perhaps I should not have said _most_.
But yes, I do believe that some don't.  No, I don't believe we should,
by default, dismiss the latter out of hand.  I think we ought to make a
best effort to make sure that things work to the extent that, _in
general_, people don't have to try to edit Makefiles to make a port
install and run correctly.

What's more telling is the fact that many problems with ports are
_not_ simple dependency problems, distfile mismatches, and the like.
There are many, much harder, problems, that we should not be anticipating
that anyone without a _high_ level of technical expertise is going to
be able to take on.  Again, GNATS is full of these kinds of problems --
especially the oldest PRs -- and more often than not, they sit there
for months and years.  I think this is A Bad Thing.

> I have
> already been bitten by the (very politely formulated) "either accept
> the ball bound to your foot or sod off" policy, I have already had
> my name written over code I heartily disagreed with (because I ended
> up as a committer-by-force on a port I send a patch for, and what
> was committed was quite a bit different from what I would even
> *consider* if it was my tree; cool, isn't it, having your name
> slapped on a lame code written by someone else, huh?)

I don't know any of the particulars you're talking about so I can't
address them.  If you don't want your name on anything, have it taken
off -- you are no more or less a volunteer than whoever committed the
code, or me, or anyone, so no one can 'force' you to do anything on this
project.  If you disagree with something that was done with your name on
it, by all means, say so.  As for what individual committers did, you
should take it up with them: if you get treatment you consider
unacceptable, then bring it to the attention of portmgr.  (If it will
allay your concerns, I will recuse myself from any followups on those
issues.)

There is obviously not going to be any agreement here between the
two camps of 'commit only things that we can guarantee will be useful'
and 'commit anything that might be useful to someone, somewhere, at
some point.'  Frankly if we go to the latter it will save me, personally,
an awful lot of effort that I try to put forth, and I will instead go
work on my own interesting hacks.

If people feel strongly enough that this single requiremet is too
burdensome, then maybe they should investigate setting up a ports-wip
(Work In Progress) repository somewhere to hold 'might work, good luck'
code.  NetBSD uses this model as pkgsrc-wip; you don't even need to be
a NetBSD committer to add things to it.  (Disclaimer: I merely lurk
on the NetBSD lists and am not intimately familiar with their processes.)

However, I should caution that a wip, or sandbox, or whatever, area,
is not a "universal solution" either: there seems to be continuing tension
between the people that want to see new packages (their term for ports)
only be committed to pkgsrc and only once they are well-tested, and
the people who just want to try out things in the sandbox area.  Then
pkgsrc users wonder why the things in wip aren't in the main tree, and
aren't satisfied with the answer "no one can guarantee yet that it will
actually work right."  Further, it divides the attention of people who
are actually interested in new packages -- which one should they try
to get it added to?

Again, I don't like the wip model and so I'm not going to spend any of
my own time working on setting it up, but I'm not going to try to talk
anyone out of it, either.  I mention it out of completeness, and to try
to offer some kind of concrete suggestion that will possibly offer a way
out of this disagreement.

mcl