[vlma-devel] TODO before releasing

[Adrien Grand]

Hi,

Since the VLMa is now almost stable, it would be very cool to release.
In order it to be usable by anyone, I think we should work around the
following features/problems before:

 1. Provide a built-in authentication mechanism.

 2. Find a way to share properly RRD graphs between the daemon and the
    web modules.

 3. Be able to run the daemon and the web interface very easily, like
    Hudson (https://hudson.dev.java.net/). 
       $ java -jar <my jar file> 
    and it just works.

 4. Save the daemon configuration file in an appropriate place. It is
    currently saved in the directory from where the daemon is launched,
    which is not a robust solution.

As I've already thought about that, here are my propositions:

1. I think lots of people would like to delegate authentication to
an existing system (their own LDAP, CAS or database) so I still hesitate
between providing a simple built-in authentication mechanism and using a
complete authentication and authorization framework (like Acegi) which
would let people configure it to fit to their environment.

2. One could save them to a predefined location (like
/var/lib/vlma/rrd) and then export them to the web thanks to a servlet
which would convert a FileInputStream to a ServletOutputStream. Another
solution would be to write the graphs directly to a web exposed
directory, but I'm not convinced by this solution because of security
reasons.

3. To work around this problem, hudson embeds a very lightweight servlet
container (Winstone, http://winstone.sourceforge.net/). I think this is
a good idea.

4. Save it to ~/.vlma/config.xml ?

Waiting for your thoughts. :-)

-- 
jpountz