[Zope] long computations

[Robin Becker]

I have a problem with a particularly cpu/memory intensive calculation.
The current server implementation is in Tcl (tclHTTPD) and uses Python
as the calculation engine in a separate process. To avoid running out of
memory I implemented the calculation as a server and just add jobs to a
server queue. The tclHTTPD implementation allows me to keep the server
busy as http requests come in and to resume pages when jobs are
finished.

I can see how to implement most of the logic in Zope2 (assuming each
request is in a separate thread), but how can I handle the case when the
request is prematurely aborted. In my existing implementation I detect
this when the http socket becomes readable and then do a cleanup to
remove the request from the queue.
-- 
Robin Becker