As a new MediaWiki developer (8 merged commits right now,
https://gerrit.wikimedia.org/r/#/dashboard/417) and an experienced
template and Gadget developer on pl.wikipedia
(https://pl.wikipedia.org/wiki/Wikipedysta:Matma_Rex):

2012/9/3 Oren Bochman <orenboch...@gmail.com>:
> 1. The community is an a massive untapped resource for development. (They
> like to edit wikis, upload photos and also to code....)

Totally this. Don't also forget about non-English wikis; there are
many JavaScript programmers and tools developers who for various
reasons do not look into MediaWiki itself. (I used to be one until
recently.)


> 2. I would seriouly look at maximizing its potential before allocating more
> funds for paid devlopment.
> 2.1 This means making it much easier to develop/test/deploy to Live wikis.
> (Short Tutorials, Code Samples, Documentation)
> 2.2 Create a culture where new coders are assigned to work with experinced
> coders to fix and maintaining existing code.
> 2.4 Team up with Wikia and WikiHow Devteams on common features and on small
> wiki testing.

All this could be great, but for me the worst "obstacle" is the
glacial pace of gerrit review. As a template/gadget developer I'm used
to a quick cycle: code, preview (a template) or test in debugger
(gadget), look at it for five more minutes to maybe catch some stupid
mistakes, and press Save (of course, I also used dev versions of
scripts or templates' sandboxes for non-trivial changes).

With gerrit, I code, check, check on my testwiki, git review (and by
god, is that tool a total kludge!), and then I wait for days for
someone to look at my changes, or head to #mediawiki and beg for
reviews (and apparently the channel is half-dead; is it supposed to be
a support channel? Because sometimes I'm the only one replying to
newcomers there...). It sometimes looks like all the experienced MW
developers are just reviewing each other's changes.

Then someone complains (sometimes about something pointless, or
something they could fix themselves in thirty seconds and submit a
patchset), I code again, git review again, and wait again.

(I don't mean this all personally, to anyone.)


>      While I applud Sumana who does a great job with the community - this
> works needs to be followed though organicaly by all members of the
> development teams

Let me just say that I basically love Sumana already, for her help and
encouragement. :)


Also. gerrit is absolute load of poop. The web UI kinda sucks (but we
all know this already, don't we?), but my real beef is with the
git-review tool. I'm semi-experienced with git, and rebases are not a
scare to me; but if it keeps complaining about multiple commits to be
sent when I only have a single new one, and if something as simple as
creating a commit that depends on two other unmerged commits requires
this much arcane magic, and if I can't post a review to go along with
my patchset, and if it requires hand-applied patches (!) to work
properly on Windows, then something is deeply wrong. I can only
imagine what kind of torture using it must be to someone just starting
out with git (or, worse, source control).

-- Matma Rex

_______________________________________________
Wikitech-l mailing list
Wikitech-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikitech-l

Reply via email to