RE: New Xalan not playing well with Tomcat 4.0?

[Joseph Kesselman]

On Wednesday, 06/05/2002 at 06:45 CST, Bryan Kearney <[EMAIL PROTECTED]> wrote:
> If the new code could cache the results.. that would be good to. In some
> profiling, getting the DTM Manager was showing up as a performance hotspot
> since it was scanning the jars.

Hm. There may be a possible issue there.

The point of the whole Context Class Loader thing (which I still don't understand as well as I'd like) seems to be that different threads may want to load from different places -- for security reasons, or because one is running an experimental version while another is running a production version, If we cache -- unless we attempt to index the cache by thread rather than using a single static field -- we might be defeating those assumptions.

I'm not sure that's a problem, but I'm not sure it isn't. I'm still trying to find a good architectural description of how Context Class Loaders are expected to behave and what we should be doing to tolerate them.

______________________________________
Joe Kesselman / IBM Research