The Incredible Shrinking Process
Some time ago I was very pleased when I found one of the causes for the Incredible Expanding Evolution. At the time I was positively pleased that the Evolution Data Server's heap usage was under 1 megabyte when in use (the benchmark was to run 20 bookviews).
Well, things have come on a long way since then:
Now it's peaking at just under 400Kb! The 200Kb chunk is the Berkeley DB cache for my addressbook, and the size of that is tunable. After that come add_module and add_alias from libc's gconv implementation of iconv. Now I'm pretty sure the data they are loading should be made constant and shared among all processes. Anybody know of existing work in this area?
NP: Fear Of Fours, Lamb