Re: GOP pre-planning: Frog bugfixing

[Graham Percival]

On Thu, Dec 11, 2008 at 10:00:41PM +0100, Valentin Villenave wrote:
> Both sides are equally important, because it's the feature count that
> makes users *and devs* stick with Lily.

New features are way sexier than bug fixes.

> > Once somebody has been doing small bugfixes for a few weeks,
> > they'll naturally start taking on bigger and bigger things.  The
> > idea behind the Frogs is to get more people doing those small
> > bugfixes.  The other stuff will take care of itself.
> 
> Nothing takes care of itself. You taught me that :)

I feel so proud!  :)

But in all seriousness: once somebody is familiar with the
lilypond internals, they're going to gravitate towards writing new
features.  And if we have a group of 10 developers, there'll be
plenty of discussion about those new features whenever somebody
gets into trouble.


> > Umm, like this?
> > http://code.google.com/p/lilypond/issues/list?q=label:Bounty
> 
> Nope, more like Jonathan's one (he's very clearly expressed why I
> think this should work big time). And in a visible place.

Earth to Valentin: there's approximately half a dozen people in
the world for whom it's remotely economically feasible to work on
bounties, and they're perfectly capable of looking in google code.

Let's take the relative \includes.  My initial guess was that it'd
be 2 or 3 hours, but Carl said it'd be longer than that.  And Carl
is one of the best "non-core" developers.  Let's say it takes him
10 hours all told -- 5 hours to understand what to do, 1 hour to
write the 20 lines of code (in total) that this feature would
require, 1 hour to test+debug, 1 hour to modify docs+write
regtest, and a total of 2 hours to send emails, upload the patch
to the online patch viewer thingie, etc.

What are people offering for this?  $100?  Wow, he's just made
$10/hr.  As a Master's student, I make more than double that
amount by teaching first-year non-CS students how to use Excel.

Making a fancy display of requested features and bounty amounts is
*not worth it*.  Getting familiar with the lilypond source code
takes a long time.  By the way, I'm including $10/hr as "remotely
economically feasible".  My guess is that if people like Carl and
Neil were to start chasing bounties, they'd be lucky to get an
hourly return of $10/hr.

Don't get me wrong; bounties are a nice idea of a bonus for
current developers.  But they basically require existing
development knowledge.  Your market is a tiny set of individuals,
not the faceless masses that contribute (hah!) to LSR or the wiki.

Cheers,
- Graham