[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Packaging Openstack for Debian: anyone else interested?



2011/3/28 Thomas Goirand <zigo@debian.org>:
> On 03/28/2011 03:18 AM, Soren Hansen wrote:
>> 2011/3/27 Thomas Goirand<zigo@debian.org>:
>> For years, Debian Developers have shouted and screamed whenever
>> someone in Ubuntu added something as benign as a patch system (so
>> that their indvidual changes would be easier to absorb back into
>> Debian), and here you are, saying you want to replace our build
>> system altogether for no good reason? No, I don't think having a
>> couple of overrides is a good reason to replace the build system, and
>> your suggesting that our debian/rules doesn't use setup.py really
>> doesn't add much technical credibility to your criticism, to be quite
>> honest.
> Calm down! It is constructive criticism.

We must have very different opinions on what constitutes constructive
criticism.

> I really appreciate the work that you have done so far, and I intend
> to reuse your work, and share my thoughts and contributions. It's just
> a shame that I couldn't chat with you on IRC over the last days, as I
> believe it would have avoid such a tone.

Since 2007, I've been logged off IRC for a total of less than 24 hours.
None of that has been in recent history. If you couldn't get onto IRC,
you could have e-mailed me (which you clearly had access to).

>> If you want to do this on your own, fine. I think it would be sad if we
>> couldn't make this good example of a "this is how Debian can be a
>> downstream of Ubuntu" story, but I can't force you to do anything. If
>> you actually expect to work with us, I suggest you stop assuming we're
>> idiots who don't know how to packages stuff and start being a courteous
>> downstream, and let's just say you've not got much of a head start on
>> that account.
> Where exactly did you see I wrote you were "idiots"? I wrote that "I
> didn't like". This show my *personal tastes*, because I like to
> understand everything that happens, and not just trust blindly the
> rules.tiny style.  Please read again...

Ok, if you don't think we're idiots, why do you go and replace our work
without even talking to us beforehand? I would expect that if you
thought our opinions were worth anything at all, you would at least have
talked to us first. If you wanted to work with us, "personal preference"
is not a good reason to diverge.

> The rules.tiny style does useless calls to dh_helper scripts, and
> makes it harder to track what's going on (you rely on the "magic" of
> the dh_auto* scripts), and also is a bit slower than it should
> (useless calls to dh_helper scripts that sometimes aren't needed, like
> for installing .desktop files and so on...), which is why I like more
> the other kind.

The *vast* majority of the build time is spent building docs or running
the test suite. You might be able to shave off 10 seconds of build time.
Let's say it takes you an hour. Let's say you make $50 an hour. You can
get 4 cores for $0.015 an hour on Rackspace Cloud Servers.

						 50 dollars/hour
----------------------------------------------------------- = ~5 million
0.015/4 dollars/cpu-hour * 10 seconds * 1/3600 hours/second

That's 5 million times you'd have to build this thing before you break
even (assuming the cost of a CPU hour stays static for eternity). Not
only that, the (now) superfluous dh_* calls you remove might be required
next week, so you're incurring a maintenance overhead that's not
currently there.

Hardware scales, developers do not. Also, I'm on a mission to build cool
technology, not a mission to save a cpu-hour spread over the next couple
of millenia.

> I would also have appreciate if we could discuss that together, I went
> on IRC to chat with you, but I had no luck with that over the last few
> days.

I'm there now. I'm "soren" (imaginatively) on IRC.

> Moreover, I hope my work wont just be "downstream", I hope we will
> constructively collaborate and work together! If what you expect me to
> do is just download your work, do a build, and upload blindly to
> Debian, that wont work, I'll be of no use.

I never expected you to, and I never said that I did. I'm puzzled,
though, what you expect to happen when you replace our work with
something else without consulting us, but say you want to "collaborate
and work together". If we're going to work together, that implies we'll
share our efforts. Are you expecting us to just accept your stuff and
blindly uploading it to Ubuntu?

> You should instead be happy that I want to improve things, and on my
> side, I hope we will both learn from each other.

I've seen nothing from you so far that indicates wanting to learn
anything. I'd be delighted to be proven wrong. Now would be a good time
to start.

>> Or, you know, you could stop building new silos and actually work
>> with us. There's no technical reason why we couldn't share our
>> packaging efforts.
> Sure, I 100% agree with that, and I intended to do so.

So why are you setting up stuff on Alioth to maintain your work instead
of working with us where we do our work?

> But, if I'm not mistaking, there are things that unfortunately *have*
> to change, like the Ubuntu package using upstart-job.

I've tried, several times, to explain to you that the dependency on
upstart-job exists *exclusively* in the binary packages, and it's put
there by dh_installinit. We don't have a hard dependency on upstart. If
Debian doesn't deal well with upstart, why does its dh_installinit let
upstart jobs take precedence over sysvinit style init scripts?

-- 
Soren Hansen
Ubuntu Developer    http://www.ubuntu.com/
OpenStack Developer http://www.openstack.org/


Reply to: