On Sun, 2009-08-23 at 21:53 -0430, Patrick O'Callaghan wrote:
Your BZ report talks about issues with large newsgroups.
Are you seeing the same with large mail folders?
I don't have any/enough mail folders that compete with the quantity/size of newsgroups.
I have one IMAP folder with over 60,000 messages in it and I don't see any sluggishness (maybe a few seconds on startup),
That's not surprising then. Not that I trust the message counts much anymore (since the sqlite migration fiasco), but the vfolder that aggregates all of my newsgroups says there are 418389 messages in it with 418344 unread (i.e. only 5 unread messages -- yeah right). However I know the difference between read and unread messages is in reality much much greater than that. Case in point for not trusting the message count numbers any more. But I digress on to yet another bug. Performance in an individual (real) folder is great. Switching from message to message is quick. It's in vfolders which represent a large aggregation where the message-to-message performance is sluggish, and of course, as I have said, startup times after a crash (which is still too frequent an occurrence unfortunately) is on the order of close to 10 minutes if not more and when I strace it, it's all reads/writes to the sqlite db(s).
but I don't use Evo for newsgroups.
You should try it sometime if you want to see the scaling problems. I use gmane and nntp for mailing lists instead of getting the mail delivered here. Doing that really highlights the scaling problems. IMHO, the evo developers (especially the ones responsible for its scalability) should be *required* to use gmane for all of their mailing lists. Nothing like having to eat your own dogfood to making it taste good. b.
Description: This is a digitally signed message part