Ted Zlatanov wrote: > I haven't seen it (which is probably my fault as I was not paying > attention), can you please resend and CC me? Sure, it's attached below. For me it makes the startup time go down from 5 minutes or more to 10-20 seconds when I start up Gnus with my ~140 imap groups over a slow mobile network. The thing is that nnimap-retrieve-groups is very sensitive to high latency. The patch doesn't remedy that, but it cuts down the number of groups it polls. > Interesting. It sounds like we should get rid of the variable in code > completely and always use an accessor function with a "decoding" > parameter (so we can get the encoded or the unencoded name). I don't > like the current hidden behavior. Do you agree? You're right, using the variable instead in situations where the encoding isn't relevant is arguably a sort of micro-optimization. It shouldn't be a significant performance hit to use the function instead. Or if it is, the right solution would be to optimize in imap.el by keeping the decoded mailbox name in a variable. I don't really see the point with adding a "decoding" parameter though. The utf-7 encoding is just for transport in the imap protocol, and I can't see any reason why higher code would be interested in that. If anything, it'd be reasonable to change imap-current-mailbox (and possibly other variables) to contain decoded mailbox names instead, but that'd of course have compatibility implications which probably don't make it worth the hassle.