Re: `viewcvs' (Was: Re: [dbootstrap] `newt' and `boxes.c', `bogl' and `bowl'[, `???' and `boxeX.c'?])
>>>>> "Joey" == Joey Hess <email@example.com> writes:
Joey> Karl M. Hegbloom wrote:
>> I'm goofing with `viewcvs'. It's a clone of `cvsweb' written in
>> Python. I think we should install it on `cvs.debian.org' in place of
>> `cvsweb', since it can do the annotation feature without needing
>> writes to "history", etc.
Joey> How on earth did they manage that?!
I honestly don't know; I've not tried to read it's code yet at all.
It's all written in Python, which I've yet to read even a tutorial
for. (I hear good things about it - C++ people adore it.)
>> It's also got a checkin database thing, where every commit is logged
>> to a MySQL database, and you can query that via a web interface... I
>> haven't looked into that much yet.
Joey> Sounds yicky. Optional, I take it.
Yes, optional. It's not very complete, I think, but potentially
useful. It could be extended to be made much more so, with language
parsers and symbol databases, etc. Try it with:
>> In order to do the transition from `cvsweb' to `viewcvs' I made some
>> Apache rewrite rules... Hmmm. I guess since it's the home page for
>> `cvs.debian.org' anyway, there's no big deal changing over. The two
>> programs are functionally identical.
Joey> Yeah, they do seem to be so.
They even look the same. It's a total clone.
Joey> I may be interested in just phasing out cvsweb from debian entirely, and
Joey> making viewcvs replace it. I'll have to take a closer look at viewcvs
Joey> first though. Cvsweb has major code cleanliness issues, which make me
Joey> not trust its security; if viewcvs has cleaner code, that alone would be
Joey> worth it, IMHO.
The code is likely to be cleaner; the README (which I imagine you've
found by now) cites that as a reason for writing it. I don't know
whether it's any more secure; as I said, I don't read Python yet, and
besides am not a CGI security expert. (I know basics, but am not a
CGI script or web guru by any means. Whew.)