I use pycurl in production on couple hundred attended and unattended devices.
No memory leaks in my use: pycurl.CurlShare(), pycurl.Curl(); python 2.4~2.7.
Important to me were:
1. openssl with its features like certs, SNI and the fact that it's
tested by millions, and
2. libcurl with all the http features and that it too is tested by millions, and
3. if something works in a browser, it works in pycurl

curlmulti is not clean in pycurl, you can make dangling references and
crashes from python.
when I tried python3 couple years back, pycurl seemed to leak memory
or references, I would think that be fixed by now, YMMV.

if you want hard timeouts, you need c-ares, then if you run on any
unix and you have very long-running processes, you need patched
c-ares, otherwise curl won't get new dns server when you move to a new
network or network switch is changed.

All in all, if you only need http/https, nowadays you can get by with
python alone.

http://blip.tv/pycon-us-videos-2009-2010-2011/pycon-2011-http-in-python-which-library-for-what-task-4896894

P.S. I'm not a libcurl/pycurl developer

On 31 July 2012 18:30, André Cruz <andre.cruz_at_co.sapo.pt> wrote:
> Hello.
>
> I'm assessing the PycURL status in order to know if I can use it for a project. I've looked at the repositories on sourceforge and Github (which one is the official now?) and both don't have commits for 2 years. Is it really that stable? :)
>
> Is anyone actively using it in production? What are the usual problems?
>
> I'm particularly worried about reports like these: https://groups.google.com/d/msg/gevent/JG-PPrmiZ2I/pa5gUdFxsLUJ
>
> Thanks for the help,
> André
>
>
> _______________________________________________
> http://cool.haxx.se/cgi-bin/mailman/listinfo/curl-and-python

_______________________________________________
http://cool.haxx.se/cgi-bin/mailman/listinfo/curl-and-python
Received on 2012-07-31