As I said, PEAR developers are welcome to get Horde CVS accounts or to just
send us patches to work on this.
Not that I wouldnt love a Horde acount :), but one of the benifits of
using pear's cvs, is that when there is something 'obviously wrong' with
a pear package - eg. A bug... anyone with a pear account can fix it - by
commiting it. (and get shouted at if they break something :)..
I generally prefer a more monolithic development style, where all
changes get filtered through one or two developers. PEAR packages
aren't ginormous pieces of software, and I don't like the sometimes
negative results of random people committing to the same block of code
all the time (e.g. gross style changes / inconsistencies, different
design methodologies, etc.).
I guess that each developers personal preference - perhaps I'm just lazy and even find it a pain to use patch :)
At present pear is configured as community free for all, where it is assumed that if you make a major change - eg. extra methods/ change behaviour.. code formating (example here being pierres fixing of the image_transform a few weeks ago) - it would be discussed first, whereas important bugfixes tend to get committed, and as long as your are reading pear-cvs, you should catch those changes..
I havent seen too many instances of these random, commits, and I think when it has happenned - and a maintainer points out - 'that is not CS compatible or not desired...' most commiters will sort it out for them....
I do wonder though as pear grows, whether pear-cvs mailing list will get overwhelmed... (or just being hopefull there :)
Is this to say that PEAR packages are inconsistent? It's true, there is
still work to do. I learn something new about making things consistent
everytime I peek at someone else's package, but it's coming along.
Are there volunteers for a janitor crew to look through the PEAR
repository and check how things stack-up on the consistency dept.? Maybe
if someone wrote a 'swingset' type application that used a good chunk of
PEAR's classes, it would be easier to spot inconsistencies.
I think it will probably comedown to opportunity - xmlrpc, while working perfectly well, is a bit of a mess as far as CS goes, next time I get round to using it, If it hasnt been tidied up I might have a go.... - but I've no idea what project might come up that would use it..
As a sort of aside, I'm also not thrilled with the general lack of
discussion in the PEAR community before changes are made to packages.
Granted, things are getting better, and I'm sure many PEAR developers
discuss the code elsewhere than this list, but it's still problematic.
I do find I get _alot_ of emails privately about my pear packages, with changes, patches, suggestions.. whatever, I do wonder if the pear-dev is a bit overwhelming for somebody just seeing one package they want to use.. (not much can be done about that really though)
Where do people discuss a lot of code changes. I know that I have my
little empire with my packages, so there's not a lot of discussion. I
can't get connected to the IRC channel (something about missing ident), so
maybe some things happen there too..
That is a pain, I had to set up bnc to relay IRC via my home machine into efnet.. - it seems it has lost alot of it's open non-ident servers recently - webgiro (gone), prison.net (doesnt allow idling)...
It would be nice if we could open a relay (bnc or whatever), that at least allowed more general access to the #pear channel
There is a bit of chat on #pear, although I seem to be asleep alot of the time (wrong timezone and all)
Regards
Alan
- brent