[SAC] Project Hosting

[Howard Butler]

On Jun 14, 2010, at 2:40 PM, <christopher.schmidt at nokia.com> <christopher.schmidt at nokia.com> wrote:
> On Jun 14, 2010, at 3:28 PM, ext Howard Butler wrote:
>> 
>> As far as precedent goes, we didn't put this condition on PostGIS when they came to us for hosting help. Or GEOS. 
> 
> I agree, and I think that is unfortunate. I'm not arguing changing
> the past, but as we grow, our hosting actually does cost real money.
> We recently had to make the transition from osgeo1 to the OSUOSL
> hosting precisely because of the growing usage of trac/SVN -- 
> load on that machine has gone down by 75% since the transition.

75% of a six-year-old, poorly optimized machine.  Not too bad, though it has been a PITA.

> 
> Free-for-all hosting is not practical, and I'd like to encourage
> us to keep that in mind.

Ok, let's apply a carrot and stick solution.  Shall we change our policy to have a carrot of hosting, and incubation/osgeo labs/your firstborn as the stick?  With github as a true, useful, and usable alternative to sourceforge, a lot of our impetus for self hosting is going away anyway.  

> 
>> An
>> incubation-worthy project that would rather just freeload system
>> resources from SAC either needs to bring with or create its
>> champion within SAC.
> 
> For the the initial setup cost of a transition, sure; though I've
> now simplified that so much that it could practically be automated.

creating new setups are not the same as importing others' trac/svn/phpcraptacularbugtracker/mailman dumps.  In the my experience, setup was more than init'ing a repository, and a big chunk of the people-time cost.  Not a cost to infrastructure at all, obviously.