On 6/13/05, Matt Wright <mw(a)mattwright.com> wrote:
I wrote the converter specifically for Wikipedia in
response to a friend's
suggestion that it would be nice and bug 235 in Bugzilla (which was clearly
written as a proposal for Wikipedia based on Village Pump discussion).
First of all, it would be nice if you could attach the patch to bug
235 so people could try it out/comment on the implementation, whether
it gets in the main tree or not.
making it an extension removes most of the
functionality (user preferences,
cache hashing, multiple language support, etc.).
Ideally extensions should be able to do anything, and theoretically
can with the $wgHooks system, if there's anything you can't do please
file a bug if there isn't one already.
It is clear that none of the top developers of the
project like the idea,
but I never really heard why. Is it because they think it can't be done well
by software or that it complicates the syntax and adds load to the servers?
(Note: Although I happen to have CVS access the following is my
opinion and not necessarily the one of the MediaWiki project, and
certainly not that of Wikimedia)
First of all, if people dislike it they can dislike it for two reasons
(that I can think of), I'll give you my view on both.
1. They don't want it in the main MediaWiki tree.
2. They don't want it on Wikimedia servers.
First of all MediaWiki is meant to be a general wiki distribution (see
meta:Sites using MediaWiki), although its most noticeable deployment
is on Wikimedia servers and indeed most development is something to
scratch a Wikimedia-site based itch we still like to keep it reusable
for other projects, the only people who seem to want this patch are
the people on the English Wikipedia who want it because they can't
agree on whether or not to use Imperial units or SI units so they
agreed on using both, and since they don't want to maintain both
versions manually or can't agree on whether to write "x km (y mi)" or
"y mi (x km)" they want to have it done automatically.
So, the % of total installations where this would be used is a very
small % of the total number of installations, very few people would
presumably want to use it outside of Wikimedia and the only projects
that seem to want it inside Wikimedia are projects written in English
which are 4-10 but there are around 800 wikis (last I checked)
Furthermore your specific implementation is designed in such a way
that people already maintaining the codebase would have more things to
maintain, and people interested in rewriting the codebase (or the
parser) would have more things to rewrite, it would touch the main
wikisyntax (hence complicating it which is always a big sacrifice),
preferences, parser cache, translation etc.
So, weighing in all the pros/cons of your implementation and the
possibility of installing it on the Wikimedia cluster not to mention
maintaining it indefinitely I think the cons vastly outweigh the pros.
Additionally, should one of the top developers note on
bug 235 in Bugzilla
that it has been discussed and is not desired, so that no one else wastes
time on this?
I don't think that's a good idea, people mainly seem to have an issue
with the idea of putting it in the main tree (your specific
implementation) rather than something like a standalone extension.