On Sep 22, 2008, at 11:18 AM, Jay A. Kreibich wrote: > On Mon, Sep 22, 2008 at 10:07:54AM -0400, D. Richard Hipp scratched > on the wall: >> I am reluctant to add to SQLite the ability to explicitly specify the >> index for a query. I agree with Alex Scotti that the whole idea >> seems >> very un-RDBMS like. > > Well it is outside of the Relational Model, that's for sure. > > Then again, the whole concept of indexes are outside of the > Relational Model.
this isn't exactly a good argument. an index surely doesn't break the relational model in any way. it's existence or absence may or may not effect execution time, but would never yield incorrect or different results. the ubiquitous b-tree index may not have been so obvious at the time, and who's to say it will continue to be a given forever. the model is just that, a model - abstracted away from implementation details, no matter how obvious they seem at the time to the implementers. to beat a dead horse, the relational model doesn't discuss anything physical at all. by your line of reasoning using disks would be outside. heaven forbid a buffer pool caching your i/o. on the other hand we have here a non standard sql extension which ties users to sqlite, and blatantly does fly in the face of the relational model. that being said, as richard points out nobody would force anyone to use this extension. i would simply pretend it didn't exist. my fear is more along the lines of what a crutch for query optimization problems features like this can become. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
