Taher Shihadeh dijo [Sat, Aug 29, 2009 at 07:38:17PM +0200]:
> > Why can this not be made part of the installation process ?
> > I mean if Cherokee fails to run unless the required changes are made 
> > then why not simply make this automatic ?!?
> >
> > Or what am I missing here ?
>
> If the config is changed automatically, the server might work, but 
> probably not as you would think.
> Silently failing is much worse than ringing all the bells and whistles, 
> in my opinion. At least this way you can debug from the beginning.

Still, we have had several configuration file upgrades taking place
behind the scenes (look at contrib/0*to0*.py), and they rarely bring
problems. If they fail to upgrade a specific configuration for a
specific corner case, all is fine (as the disruption will hopefully be
minor and easier to fix).

Not providing a way for a working service to keep working after a
minor update, and even not mentioning the need to do this update (with
full, precise instructions) in the release announcement¹, speaks bad
about the project's stability and fitness for use in production
servers. 

Given that the uptake of Cherokee installed among Debian users seems
to be quite low², specially compared with similarly oriented packages
(small, efficient webservers, such as nginx³, lighttpd⁴, boa⁵,
thttpd⁶ — There are more, yes, but they mostly seem like an excercise
on programming rather than attempts at a full-blown webserver), I
think the following question should be raised: Does Cherokee, in its
current form, benefit in any way of being packaged for Debian? 

Given Debian's long stabilization cycles, we are currently shipping
0.7.2-4 in our stable release, and are still committed to supporting
0.5.5 for the previous release at least until February. It is no
secret I am just unable to provide much support in case it is needed
for Cherokee proper (I can try to fix packaging issues and patch
minor, well-detected situations, but would not attempt to backport a
fix to a two year old release). And that situation will probably hold
for the future — Users are always advised to update to the newest
versions when they request any kind of help; there have been several
regressions in the last couple of months, and I'd really prefer
sticking to 0.99.20 as it seems to me to be way more solid than newer
versions.

So, to make it short: Should I continue to package Cherokee? Or would
you consider better for Cherokee's current development stage to be
available straight from you? 

Of course, you are free to use my packaging (and I'll continue trying
to find some free time to upgrade to the newest recommended practice),
but I think the best solution right now will be to drop Cherokee from
Debian, and from all other distributions with a long release (and
maintenance) cycle, and focus on up-to-date packaging provided
straight from you.

¹ http://lists.octality.com/pipermail/cherokee/2009-August/011036.html

² http://qa.debian.org/popcon-graph.php?packages=cherokee

³ http://qa.debian.org/popcon-graph.php?packages=nginxhttp://qa.debian.org/popcon-graph.php?packages=lighttpdhttp://qa.debian.org/popcon-graph.php?packages=boahttp://qa.debian.org/popcon-graph.php?packages=thttpd

-- 
Gunnar Wolf • [email protected] • (+52-55)5623-0154 / 1451-2244

Attachment: pgpSyW7PlUfLd.pgp
Description: PGP signature

_______________________________________________
Cherokee mailing list
[email protected]
http://lists.octality.com/listinfo/cherokee

Reply via email to