"Brian Beck" <[EMAIL PROTECTED]> writes: > Ah, got'cha. It seems impractical performance-wise since there is > indirection through HTTP, and it will take some extra work to write > complex queries involving both tags and separate attributes. But I was > actually impressed with the responsiveness on thraxil.org, so I'll give > it another try until it works. Thanks for clearing that up,
the performance hit really isn't too bad. on my modest home machine, making a fairly complex query to tasty takes 13ms (2ms on our slightly less modest server at work). the HTTP overhead involved in calling a service on the same machine (or over a very fast local network) isn't much more than making a database query. when i wrote the engine for thraxil.org, i coded it up in the most straightforward, easy to understand manner i could and then profiled it a bit (cherrypy has some nice profiler integration) and identified a couple hot spots where i just hooked in memcached with a couple lines of code. caching is your friend. worrying about performance without actually benchmarking or profiling is usually a waste of time. -- anders pearson : http://www.columbia.edu/~anders/ C C N M T L : http://www.ccnmtl.columbia.edu/ weblog : http://thraxil.org/
pgpRK6n5Pxut0.pgp
Description: PGP signature
