On Wed, 2003-06-18 at 17:06, Paolo Molaro wrote: > and will continue [t]o be static Really? Thats not my plan! I would like to see the new format be fast and compact enough that we could do this dynamically (with a bit of caching) on the server. I am also thinking about giving the XML + XSLT to uplevel clients (moz and ie6), as it would take much less bandwidth and allow for us to do some other cool things. > so if he has ideas to speedup the status pages, he can apply them to the current > stuff. The things I would suggest involve changing the format of the xml that we output not the xslt style sheet. And I have no interest in fixing up a tool that is about to sink ;-). > Besides, if Ben has suggestions, he can well make them on the list. Well, basically, my suggestion is to make it so that the output of tool can be parsed in one forward only pass. One of the problems with XSLT and XML is that it is *very* easy to program something that requires the transformer to double back. Right now, I think ours suffers from this. Basically, the only types of select queries that should be needed to directly get values should be ones of on the attribute of the current node. For example this would be a BAD format: <class name="blah"> <method name="blah" implemented="false" /> </class>
Because to find out if the blah class was fully implemented, all of the children would have to be scanned. We don't make errors that are quite as bad as that, but if you look at the xml, there are still similar constructs. _______________________________________________ Mono-list maillist - [EMAIL PROTECTED] http://lists.ximian.com/mailman/listinfo/mono-list