On Wed, Dec 17, 2008 at 09:29:41AM -0800, Stephen Hahn wrote: > So, Tom's correct. Unfortunately, that means that this choice isn't > acceptable without sizeable wins, because it has a significant impact > on minimization.
Agreed. I'd like to mention that Nico has been actively voicing support for having some sort of a sqlite3 backend (I think as a cache of the raw manifests) which could be used to get fast access to the actions, and help compute image plans faster, etc. I don't know what the truth to truthiness ratio is for that, but it may be worth exploring. That path would also have impact on search, obviously. I assume it has a great deal to do with the database schema, but it might be possible to rely entirely on sqlite for search as well. I don't know the pros and cons of such a move, but it might be worth looking into. It's another expansion of the requirements, but sqlite is smaller than a JRE, and it's already required for SMF (albeit SMF has, at present, a private copy). And it wouldn't strictly be required for core functionality (or even all functionality), it would just be there to speed things up. Danek _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
