On Tue, 15 Jun 2004, Seth LaForge wrote: >Well, in my case, the cache file is only 1.1MB, which I would expect to >take only about 22ms to write, if it were actually going to disk rather >than cache. This leaves about 2.480 seconds unaccounted for. I'm >guessing this must be the scan you describe. Have you put any effort >into speeding that up for 1.3? If not, maybe I'll take a look at it...
I'm open for suggestions. But just for the sake of testing, you could try to remove all calls to fsync() from Binc, then compile and see if you notice any difference. I'm not quite sure where the 22ms comes from, did you measure this somehow? >> The good news is that in 1.3, we'll add the MULTIAPPEND extension >> (rfc3502.txt). This allows clients to copy several messages in one go, and >> Binc will then delegate several UIDs in one go, and the cache file will >> only be written to once for the entire copy operation. >Do many clients actually support MULTIAPPEND? If not, it doesn't help >much... All sane clients should, but I must admit that I don't know which do and which don't. Andy :-) -- Andreas Aardal Hanssen | http://www.andreas.hanssen.name/gpg Author of Binc IMAP | "It is better not to do something http://www.bincimap.org/ | than to do it poorly."
