GAE and scaling
now that GAE is days old, lots of wisdom is appearing on 'the internets.' and some of it quite interesting. this post is a nice summary so far.
much of the talk is about the slowness of GEA data calls. another issue is the lack of session state and the perceived limits of Python. it all boils down to scaling and targets. to me, this is a peek into the thinking in google-land. this engine is all about making web responses lightweight and super fast. and that means limiting what you can do in a response (no fancy computing of totals, averages, etc.) and what you can carry around in response metadata (no session state to resolve and drag about, etc.).
i've thought about this quite a bit when building exyus. i *do* use SQL databases (via XSLT transforms) and i *do not* use session state. i decided to limit the work to classes that are tuned to HTTP methods; to make media-types a first-class feature of the framework and to limit the work done to responding to HTTP methods.
ok, sure - my framework doesn't scale like GAE - but hey, it scales just nicely for me!