Forum OpenACS Development: Response to rp_serve_abstract_file and PerformanceModeP

Collapse
Posted by Don Baccus on
Threads aren't deallocated at the end of each connection, no.  You can control how long threads live via parameters in your nsd.tcl init file.

ns_share's deprecated which is why you see nsv_* being used all over the place.  It's been ages since I read the justification for the switch in the AOLserver docs so I don't remember exactly why, though nsv_* are cool so I don't really care.

What you're suggesting is that the map be cached across all threads rather than have each thread cache their own copy.  In theory this sounds wonderful but given the time it took us to figure out why the existing aD code you copied failed and the easiest way to do internal redirection properly, switching to system-wide caching of the map makes me a bit nervous.  There might be hidden gotchas in there we don't know about.

But if you or anyone else has time to look into it with an eye towards convincing us that it would be safe, yeah, why not?

Anyway ... I'd love to see your internal redirect get into the CVS tree, and for the various abusers of rp_serve_abstract_file to get changed to use it, ASAP.  Let me know when you're done and we can talk about making the change.