incubator-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Meeks <>
Subject Re: Blondie's Parallel Lines... numerically ...
Date Fri, 03 Jun 2011 00:57:27 GMT

On Thu, 2011-06-02 at 16:55 -0400, wrote:
> Questionable?  If only 54 people have checked in code in the last 6 
> months, then no amount of magic with source code indentation is going to 
> get you to 400 developers.  If you disagree, I'd like to see the magic you 
> can do with the tab key!

	While I agree that this is increasingly irrelevant; there is an
important issue of competence and accuracy raised here that may be
generally interesting; I would point out that Rob's contention:

On Thu, 2011-06-02 at 15:34 -0400, wrote:
> When I check the commit logs for LibreOffice and apply the Apache
> criteria for what defines an "active" participant (a commit within
> the last 6 months), I see only 54 names.

	Is almost certainly based on a basic misunderstand of how the
LibreOffice code is structured - across multiple git repositories. If
you count just 'bootstrap' (that includes some, dmake and
some scripts[1]) you might get that - but that is for a tiny fraction of
the code. Assuming your definition of 'active' is indeed that any commit
within the last six months qualifies - then we track those numbers too.

	To me it is sad to, so confidently, propagate these clearly
contradictory numbers - without first privately flagging an error in our
published code for counting the stats that Cedric[2] maintains and is
based on Jonathan Corbett's excellent gitdm. Perhaps you found one ?
Clearly we don't pull these numbers from a hat weekly :-)

	While it is only peripherally relevant to the matter at hand, the stats
I generated, after applying a mini patch to our 'log-data':

-    $scripts_dir/merge-log -p LIBREOFFICE_CREATE.. >$outdir/all-lo.log
+    $scripts_dir/merge-log --all --since='2011-01-03' >$outdir/all-lo.log

	Show 'active' contributors by affiliation - ie. at least one patch
contributed in the last six months like this:

Employers with the most hackers (total 214)
(Unknown)                  138 (64.5%)
Oracle                      45 (21.0%)
Novell                      18 (8.4%)
Known contributors           7 (3.3%)
Canonical                    4 (1.9%)
Redhat                       2 (0.9%)

	And in the last ~9 months (since LIBREOFFICE_CREATE) - our more normal
datum - we have:

Employers with the most hackers (total 283)
(Unknown)                  194 (68.6%)
Oracle                      54 (19.1%)
Novell                      19 (6.7%)
Known contributors          10 (3.5%)
Canonical                    4 (1.4%)
Redhat                       2 (0.7%)

	Which we try to graph month by month (clearly far fewer are active each
month). The 200 number we hear is an approximation of the "entirely new
contributors with committed code since we started". The 'Known
contributors' are mostly volunteers too. 200 seems a good enough, round
representation of 283 hackers[3] for me :-)

	Casting aspersions on the value and size of volunteers input I fully
expect (personally I value lots of diverse small changes from
contributors myself), but the quantification I didn't.

	Our gitdm setup, affiliation database and patches are here:

	Perhaps there are indeed some factor-of-four bugs in it, if so we'd
love to know, it is our intention to print reasonably accurate stats.
Patches appreciated, this is Free Software right ;-)

	All the best,


[1] - hint: ./ and run download to get the rest
[2] -
[3] - this should not include the ~200 translators helping out TDF
      either; we don't count .sdf or .po files
--  <><, Pseudo Engineer, itinerant idiot

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message