Merge lp:~toshio/loggerhead/mod_wsgi into lp:loggerhead

Toshio,

Thanks for this. I wonder whether we ought to put the various files into a contrib/mod_wsgi directory instead of putting them into the project root. What do you think?

Some minor doc tweaks:

* loggehead -> Loggerhead :-)
* "bazaar conf" -> bazaar.conf
* "Loggerhead's doc directory" assumes installation via a package manager doesn't it?

On the legal side, loggerhead.wsgi needs a Copyright 2010 Canonical line added above the GPL stuff. Could you please also sign the Canonical Contributor Agreement? See http://www.canonical.com/contributors.

Forgive my ignorance but how does the MANIFEST.in file come into play?

toshio wrote on the mailing list that the wsgi script was originally written by Stuart Colville:

http://blog.projectfondue.com/2009/11/26/loggerhead-and-mod-wsgi

So we might need him to sign the contributor agreement too.

> Thanks for this. I wonder whether we ought to put the various files into a
> contrib/mod_wsgi directory instead of putting them into the project root. What
> do you think?
>
Depends. The script is installed by setup.py so I'd lean towards putting it into the root just like serve-branches. However, if you don't want using mod_wsgi to be a supported method of running loggerhead, then I could see not installing it to /usr/bin via setup.py and moving it into a contrib directory. The example config files follow the same convention as the loggerhead.conf.example file which is in the project root currently.

> Some minor doc tweaks:
>
> * loggehead -> Loggerhead :-)
> * "bazaar conf" -> bazaar.conf

Fixed. Branch updated.

> * "Loggerhead's doc directory" assumes installation via a package manager
> doesn't it?
>
Nope. Try this::

  python setup.py install --root /var/tmp/loggerhead
  cd /var/tmp/loggerhead/usr/share/doc/loggerhead/

A package manager would likely included a version number in that final directory, though, which is why I say "Loggerhead's doc directory" instead of being more specific.

> On the legal side, loggerhead.wsgi needs a Copyright 2010 Canonical line added
> above the GPL stuff. Could you please also sign the Canonical Contributor
> Agreement? See http://www.canonical.com/contributors.
>
I'll need to send the agreement on to my manager but I don't see any difficulties with it; I know other people who have been able to sign it in order to contribute to upstart. However, I can't, in good conscience add the header by myself. As noted in the linked bug:
  https://bugs.edge.launchpad.net/loggerhead/+bug/566118

Stuart Colville (username: muffinresearch) originated the vast majority of the script. I've only done a few tweaks to make it configurable via a config file. I've emailed with him and he's fine with signing the Canonical Contributor Agreement (and also works for Canonical now.) How should we work out that aspect? Does he need to sign the agreement and then give me a copy of the file that has the header? Can I allow him to push to my branch somehow? Can you add the header after Stuart and I both sign the agreement? I just need to know what needs to be done to make the assignment clear; I think both Stuart and I are anxious to see the code committed :-)

> Forgive my ignorance but how does the MANIFEST.in file come into play?

When you use setup.py (distutils or setuptools driven), you get a command::

  python setup.py sdist

That command creates a tarball of the project. Scripts and python packages listed in the setup.py file do not have to be listed in MANIFEST.in but other files (non-standard doc files, for example) do otherwise they are not included in the tarball.

I've sent in mu Contributor Agreement. Let me know what we need to do to get Stuart's code all legal and kosher as well.

Ian, I've been added to contributor-agreement-canonical now -- let us know what we need to do so Stuart's portion is taken care of as well.

hi Max, I'm still interested in this. I don't know whether I need to rebase to trunk or anyhting but I'm willing to do that work if necessary once we get the legal situation wrt Stuart figured out.

Okay, I will find out what needs to happen there with Stuart.

I'm happy to place my contribution under loggerhead's license as per my current agreement as a Canonical Employee. Let me know if there's anything further you need.

Thanks for updating this, Toshio! Sorry that it took me so long to get to it.

I really love all of the documentation and examples, they all look great.

However, in loggerhead.wsgi, instead of having a bunch of custom code, we should re-use loggerhead.main as much as possible. See what serve-branches does, for example.

Otherwise, we're going to be disabling a lot of features when we use WSGI, and we'll also add another main entry point that we have to maintain, which I would really like to avoid.

Oh and also, yeah, you'll have to rebase to trunk since a few things have changed, mostly in setup.py (because start-loggerhead and stop-loggerhead no longer exist).

Also, there's already this WSGI block in loggerhead.main:

    else:
        if protocol == 'fcgi':
            from flup.server.fcgi import WSGIServer
        elif protocol == 'scgi':
            from flup.server.scgi import WSGIServer
        elif protocol == 'ajp':
            from flup.server.ajp import WSGIServer
        else:
            print 'Unknown protocol: %s.' % (protocol)
            sys.exit(1)

Could we perhaps just modify that to get direct mod_wsgi support?

I'm going to move this back to Work In Progress since it's been sat in the active reviews queue for a while.

This can be tweaked by a reviewer and landed, its this || close.

I've done so. I think there is room to cleanup the tree, move stuff out of root etc, but this should be sufficiently together to work. Max's suggestion about loggerhead.main would be good too, if someone wants to follow up.