Re: Octave-Forge bugs in the tracker?

[Thomas Weber]

On Sat, Jul 23, 2011 at 01:45:37PM -0500, Jordi Gutiérrez Hermoso wrote:
> On 23 July 2011 13:13, c. <address@hidden> wrote:
> 
> > It seems the majority of those who spoke on this topic have an
> > opposite point of view, but actually I'm not a great fan of having
> > octave-forge and octave too close together, I think they are two
> > different projects and that it would be convenient to keep it so.
> 
> Our users disagree with this all the time. And most of the time, so do
> we. 

Who's "we"?

For the record, I disagree with a lot of things that people here seem to
take for granted:
1) I do not think that there is really that much overlap between -forge
developers and core developers. 
2) If my feeling in 1) is true, than forcing a move to a different VCS
onto those developers is plain wrong. I do not know care a second how
much hypothetical advantages there might be: if the -forge developers
don't come forward themselves with the wish to move, than everybody
outside should just stop pushing them - and yes, that includes -core
developers!

So, back to 1). Using 'svn log' and snipping the whole log file with
vim, I created a log of svn commits into octave-forge of the commits of
2011, in the file svn-stats-history-2011.log.

Next step: extracting the committers:
$ cat svn-stats-history-2011.log | grep '^r' | awk '{print $3}' | sort -u
abarth93
adb014
and
asnelt
blondandy
carandraug
cdemills
cdf
dreisamrd
gianvito
hauberg
i7tiol
jgpallero
lindnerb
lmarkov
martinhepperle
paramaniac
prnienhuis
przykry2004
rafavzqz
schloegl
thomas-weber
to
treichl
with

There are some obvious false words in there, maybe due to some log
messages. If anybody feels like it, pipe the output of the 'grep' part
into a file and remove the false stuff. 

So, something similar for the octave repo at Savannah:
$ hg log --date "2011-01-01 to 2011-08-01" | grep user | sort -u | cut -d' ' 
-f2-
Alois Schl?gl
Andriy Shinkarchuck
Ben Abbott 
Benjamin Lindner
Carlo de Falco
Carn? Draug
Daniel Kraft
David Bateman
David Bateman
David Finkel
David Grundberg
David Wells
Giles Anderson
Iain Murray 
Jarno Rajahalme
Jaroslav Hajek
John Bradshaw
John Eaton 
John W. Eaton 
John W. Eaton
Jordi Guti?rrez Hermoso
Kai Habel
Karsten Trulsen
Konstantinos Poulios
Konstantinos Poulios
Lukas Reichlin
Marco Atzeri 
Marco Caliari
Michael Creel
Michael Godfrey
Olaf Till
Pascal Dupuis
Patrick H?cker  <magicmuscleman>
Paul Boven
Petr Mikulik
Philip Nienhuis
Rik
Robert T. Short 
Sean Young
Tatsuro MATSUOKA
Thomas Weber
Thorsten Meyer

I've removed the mail addresses - people appearing more than once had a
different mail address. I don't know all usernames of the -forge committers,
but I'd say we have at most 50% overlap.

And as someone on both lists: I don't think the overhead of moving the
repository to git, Mercurial or whatever is worth the effort!
octave-forge development is pretty centralized right now and that is an
advantage for the current type of developers and users we are
approaching.

> When they say, "I want to accomplish such and such task", how
> often are we able to not mention some Octave-Forge package? The
> separation between Octave and Octave-Forge is awkward and is
> unnecessarily segregating our very small community.

Moving some functionality out of Octave and into packages was a
conscious decision. Joining them 'a little bit' again seems far worse
than either just one project or two really separated projects.
Just look at the awkwardness of committer rights you immediately get.

> > For example many of those who contribute to octave-forge or that
> > contributed in the past would not have done so if requirements for
> > acessing octave-forge would not have been much less strict than
> > those for octave (in terms of coding standards, licensing terms
> > etc.)
> 
> If Octave-Forge were hosted in Savannah, there is no requirement to
> follow GNU coding standards, and any license is fine as long as it's
> GPL-compatible, which Octave-Forge packages have to be anyways except
> for the non-free ones nobody uses.

We are talking about nongnu.org here, right? From
https://savannah.gnu.org/:
"Savannah aims to be a central point for development, maintenance and
distribution of official GNU software. In addition, for projects that
support free software but are not part of GNU, we provide
savannah.nongnu.org."
 
> > One more thing to consider is that, especially taking into account
> > the VERY limited work force taking care of octave-forge at the
> > moment, I would suggest to think twice before making "revolutionary"
> > changes that would require a non negligible amount of work. OF is a
> > quite successful project that, thanks to the great hard work done by
> > many in the past, currently gets on with very little effort by very
> > few people (mostly Søren only). Sure, there is lots of ways OF can
> > and should be improved, but past attempts to throw it all away and
> > start over have not gone very far
> 
> The amount of work to do here is very limited. Moving from SourceForge
> to Savannah is basically just re-uploading the code and updating a few
> urls in the web pages. I don't see a reason to move the web hosting.

If you don't move everything, you have two sides to somehow keep in
sync, including user accounts for the administrators and whatever.
How do I announce a new release from Savannah on the web pages on SF,
for example? 

> > Finally I am not sure that having a single bug tracker for all
> > octave-forge packages makes any sense. The current approach is to
> > let each package maintainer take care of his/her own packages and be
> > responsible for their status. Often when users report bugs in the
> > mailing list, they are instructed to contact the package maintainer
> > directly. I even doubt many package maintainers would ever look into
> > the common bug tracker.
> 
> This only works for the packages that actually have a dedicated
> maintainer, which most don't. Thomas Weber recently droped a bunch of
> Octave-Forge Debian packages because they were unmaintained.

Unmaintained packages are unmaintained. Neither the used bug tracker,
the used VCS nor the web hosting changes anything about that. Technology
doesn't change the schedule of people.

In fact, having a single repository for each -forge package - have you
guys looked at octave-forge recently? We are talking about 80-100
repositories here. I mean, it is actually trivial to set up a repository
at bitbucket, github, ... and develop a package there. Is there any
real, hard evidence that more people would join the project if the VCS
hosting and VCS system used would change? And more, what is hindering
those people to use one of the gazillion different hosting sides right
now?

> We can always tell Savannah to CC a particular maintainer when a bug
> discussion starts, the maintainer can then go to Savannnah and follow
> that bug there. 

Which means more work for the maintainer, who has to follow-up on
Savannah.

> As for the rest, we can use a collaborative maintenance approach where
> we have a common publicly available list of problems so that anyone
> can go to that list and take care of the problems. 

Which means more work for the maintainer, who has to follow-up on
Savannah.

> There are a number of Octave-Forge packages I would like to patch
> myself, but I find Octave-Forge unwieldy to work with right now.

Oh, come on. Checking out SVN and patching it is not that much work. I
do it for the things I find in Debian and I usually don't even care
about the packages.

        Thomas