Hi people,
As a message of general interest:
I had an off-list conversation with Daniel Marchese, a fop-user who
profiled a run of a document containing one large page-sequence of
+400 pages with a lot of tables. He showed heap snapshots of FOP
0.20.5 on several JVMs, and it always needed at least 1GB to complete
the job. He shared the source document with me, and I ran some tests
one the current trunk, gradually increasing the max-memory until it
passed, and it ultimately needed 'only' 768MB.
Still a lot, but a 25% decrease for identical input is definitely
worth mentioning, so...
Cheers
Andreas
