on 6/22/03 3:32 PM Christopher Oliver wrote: > In Linotype I saw some other strange uses of the cocoon:// protocol that > appear to be very bad for performance: > There's a pipeline that uses the directory generator to read a > directory, and transforms it using xslt into a cinclude template. > Another pipeline uses this as a source for its generator using the > cocoon:// protocol. This would seem to defeat any possibility of > compiling the cinclude template (or am I missing something?)
No, you are totally right: linotype is *very* inefficient from an algorithmical perspective. In fact, linotype does need a database if we want it to scale. I'm planning to keep an eye on JSR-170 (Repository API) and the Slide implementation for it. right now, linotype is just a proof of concept to show how things can be done quick and dirty and still be rather powerful and useable. Also to show how powerful a few tens of Kb of javascript code spread across teh client and server can be. Also, it's a try to test the darwinistic development principles of "start with the little design as possible": seed first, incrementally improve the system. Will it work? I don't know, but it's worth trying, expecially in a great community like Cocoon's, there is a lot of ecosystem power and plenty of social noise to feed the system and see what happens. -- Stefano.
