Re: Couchdb package

[Raphael Hertzog <raphael@ouaza.com>]

Le samedi 09 janvier 2010, Sam Bisbee a écrit :
> On Sat, Jan 09, 2010 at 06:19:27PM +0300, Sergei Golovan wrote:
> > On Sat, Jan 9, 2010 at 5:36 PM, Raphael Hertzog <raphael@ouaza.com> wrote:
> > > Le samedi 09 janvier 2010, Sergei Golovan a écrit :
> > >> +    - split package into couchdb and couchdb-bin binary packages, so that
> > >> +      the programs can be installed without starting the system service
> > >>
> > >> We had a discussion on splitting couchdb and decided to keep a single package.
> 
> Raphael, for your reference this is the discussion thread:
> http://lists.alioth.debian.org/pipermail/pkg-erlang-devel/2009-November/000048.html

Thanks for the pointer. But I'm with Ellioth, your arguments are not convincing.
Splitting a source package in two binary packages has never been
forking... and using per-user couchdb instances is supported by upstream since
couchdb has no code changes applied in Ubuntu to make this a reality.

David Paleino (cced) is going to package desktopcouch for Debian and he
will have to solve the same problem...

You told several times that you should get the opinion of the upstream
developers, I'm not sure it's required, but if you make it a requirement
it would be nice that either you or Elliot go on and contact upstream
to see if they believe that desktopcouch is a correct use of their software.
And if they say yes, then please support it properly.

Other solution, if you have doubts about the cleanliness of the split, I
suggest you ask advice from the Debian technical committee. I'm pretty sure
they would endorse the change. I'm ccing them to have their opinion.

Quoting some points of the former discussion:
>> Why can't desktopcouch use a system-wide couchdb service?
>
>I think his point is around the need to start couchdb at boot if you want to do
>on demand, but we've covered that.

I think you haven't. Starting a system-wide service from user session means
getting root rights in one way or another, I think it's way better for a service
that can run entirely in the user-session to run that way, we don't need any root
privileges from what I can tell (except the fact that we have to switch to
the couchdb user in the system-wide installation).

In the end, Sam, your arguments about useless splitting of the package are really
not convincing to me. The reasons for the split have been explained at length
and I think they are valid.

Please reconsider the decision or provide an alternative solution that does not
involve the local admin/user having to manually change the default
configuration to disable the system-wide daemon.

> Thanks for sending your thoughts Raphael. The init script is on the top of my
> TODO list right now, but I've also added a good number of the things that you
> forwarded along.

Cool and thanks for your work!

Cheers,
-- 
Raphaël Hertzog -+- http://www.ouaza.com

Freexian : des développeurs Debian au service des entreprises
http://www.freexian.com