[Live-demo] Alternative OSGeoLive build process

[Hamish]

Hamish wrote:
> > today I'll abstract the USER_NAME to main.sh in a fail-safe way (in all
> > scripts). there may be a minor problem here or there but overall it
> > should not be dangerous. I'll do it in a way so that scripts will
> > still be able to run independently, by defaulting to "user" if it
> >isn't set.

Cameron:
> This sounds good if I'm understanding you correctly.
> We should aim for each script to be able to run inside any
> ubuntu distribution, independent of other osgeo-live scripts
> needing to be run beforehand.

ok, it's done now, hopefully free of typos. :-) If USER_HOME is set
(and `export`ed) before you run a script it should respect that. One
positive side to this is that you can now run the install scripts to be
targeted at your own user acc't on your home system if you like.

The "user" name is set/hardcoded in main.sh.

The main trouble spot I notice is with the quickstart docs and the
application rc files in app-conf and app-data. I didn't touch those.

For the docs I'm guessing that most intelligent humans can translate
the intention on the fly, for the rc files I'm not sure, if they can
be altered to refer to /usr/share/data instead of the /home/user/data/
symlink maybe they can be generalized too.


wrt "independent of other osgeo-live scripts needing to be run beforehand"
I'm not sure; I think we can say "you can assume java is already installed,
you can assume apache is already installed, ..." rather that repeat that
setup a dozen times, each time slightly differently.  Perhaps Postgres/GIS
+ imported geodata is the most important out-of-script dependency to
consider.

But yes, you are right, the fewer prior assumptions that the scripts have
to rely on, the better!


Hamish