[Top][All Lists]

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

our development mess

From: Graham Percival
Subject: our development mess
Date: Mon, 3 May 2010 17:03:26 +0100
User-agent: Mutt/1.5.18 (2008-05-17)

On Mon, May 03, 2010 at 05:16:09PM +0200, Werner LEMBERG wrote:
> Please, PLEASE!  All parties, calm down! Meanwhile the list members
> should know how to tackle David; basically, I don't see hostility in
> his replies but rather the wish to improve lilypond.

Agreed on both points.

Look, lilypond development is a mess:
- we have way too few people reviewing patches; people are lucky
  if somebody glances at it within a week.
- we have a mixture of programming languages.
- the code base is complicated, with many parts that even Han-Wen
  and Jan say are garbage.
- much of the code isn't documented.
- we don't even agree about code formatting (in various
  languages), and/or the code formatting isn't documented.  I
  suspect that a portion of patches that are accepted only get in
  because we don't notice the formatting problems.
- contributors understandably squeamish about trying to review
  patches in areas they haven't looked at, and even more squeamish
  about trying to fix bugs in those areas.
- we'll be passing 400 open issues soon, with no indication that
  we'll soon start closing more issues than opening them.  In short,
  we need a loan from the IMF to get our issue-debt+deficit under
    (err, if people read this in the archives in a few months
     and don't get it, I'm making a joke about the Greece
     financial meltdown in April 2010)
- the build system is horribly out of date and consists of more
  duct tape than original material.
- we had a thousand-hour Grand Documentation Project to train
  people to handle docs so I could work on other stuff, but a
  year after GDP,  I was back to being the main doc person.

On the other hand, we're better off than we were a year ago:
- due to the Frogs, we have a growing number of people reviewing
  patches and working on issues.  I said *growing*, not *growed*.
  We still have a critical shortage of really knowledgeable
  developers (and/or a shortage of time/energy from knowledgeable
  developers), but at least we're moving in the right direction.
- new issues sometimes get patches fairly quickly; our "issue
  deficit" is going down.  Even better is that the higher-priority
  issues are getting more attention.
- we have regular releases.
- the two new doc editors show no indication of wanting to run off
  and learn scheme and work on new features, so I'm hopeful that
  by the end of the year I can start working on code bugs.

The question is now "what's the best way to improve things"?

There's a lot of things we could try.  Many people want to talk
about syntax and input file formatting; we could start GLISS.
There's a lot of jobs that normal users could handle to free up
developer time for other tasks; we could start GOP to recruit
them.  The build system needs a complete overhaul; we could start
the 100+ hour task of writing a new build system.  We could start
modifying/writing a python or scheme program to handle source code
formatting so that at the very least we can stop wasting time
arguing about that stuff.  We could try to think of ways to
encourage more people to review patches... maybe turn it into a
game, maybe make a concerted effort to encourage *everybody* to
review patches and ask any questions, even silly ones, and be
super-polite to any scheme newbies who ask stupid questions about
a patch, etc... etc.

At the moment, though, I think the best way to improve things is
to put all that on hold and get 2.14 out the door.  Limp along
with the current system, with all its faults, for another few
weeks.  Once 2.14 is out (well, maybe 2.14.1) and things seem
settled down, *then* we can start a big discussion about project

- Graham

reply via email to

[Prev in Thread] Current Thread [Next in Thread]