develooper Front page | perl.perl5.porters | Postings from August 2013

Re: the GitHub perl mirror

Thread Previous | Thread Next
Nicholas Clark
August 19, 2013 10:41
Re: the GitHub perl mirror
Message ID:
On Mon, Aug 12, 2013 at 06:19:09PM +0300, Leon Timmermans wrote:
> On Mon, Aug 12, 2013 at 3:20 PM, Ævar Arnfjörð Bjarmason
> <> wrote:
> > There's a lot of incoming links to that repo (I use it as my main way
> > for linking to perl5.git commits, amongst others), and it's got ~130
> > forks and ~450 watchers on GitHub.
> >
> > It's had a grant total of 13 pull requests over 2 years. I think the
> > value of it being there as-is even with pull requests enabled is *much
> > greater* for giving perl exposure, not breaking people's watch lists &
> > links than breaking all of that by having them remove the mirror.

If exposure translated into unsolicited third party contributions I'd buy
into this. But I don't think that it has.

As best I can figure out from looking at the commit history over the past 10

1) it's very noisy
2) there is no correlation between commit rate or contributor rate and
   the switch to git
3) there is no correlation between commit rate or contributor rate and
   the move to monthly dev releases, or to annual releases

I don't know when github started offering a mirror, but I doubt that it's
possible to find any sort of evidence that it helped get more contributions.
It might have, but it will be too small to measure.

I'm pretty sure that the problem isn't exposure. It's that the codebase is
written in a language that most Perl programmers don't know, that it's old
and hoary, and that isn't possible to change without being self-defeating.
In that, it's very hard to simplify the code without introducing regressions
or breaking something that worked. Do that too often and firms and OS
distributions don't upgrade. At which point you have a lovely language that
no-one uses.

In Kiev, mst gave the example of TCL 8.6, which introduces a MOP. But he said
that it's not quite compatible, hence Debian is still on 8.4. Similarly,
Mercurial states that it has no plan to move off Python 2:

> I actually use it myself with some regularity when doing archeology.
> It tends to provide a much more intuitive interface for than than the
> command line tools (I figured out how to do that eventually, but git
> is inconsistent enough that I have to trial-and-error it half of the
> time).

I find the the gitweb interface running on most useful
for this. I have a blame: search set up in Chrome which jumps straight to
the blame view of any file.

Nicholas Clark

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About