From: Matthias A. <mat...@gm...> - 2006-07-12 18:15:18
|
On Wed, 12 Jul 2006, Uli Zappe wrote: > Am 12.07.2006 um 08:05 schrieb Matthias Andree: > > >And I even considered warning of deamon intervals shorter than 5 > >minutes. In fact, with many sites, polling in shorter intervals > >just earns you a lockout. > > I know this behavior from the pop mail server of my university. > However, the critical interval that triggers a lockout after a few > trials is 10 seconds (which sounds reasonable to me). GMX are counting connection attempts per unit of time, without stating clearly how much is too much. Every 5 minutes works. web.de limit poll intervals for non-paying users for 15 minutes and 1 minute for paying users (but given their abysmal performance and contact ways, I'd rather not subscribe to their premium service). > >Trying short poll intervals with "keep" doesn't work too well either. > > What could possibly go wrong? The traffic ramps up if considerably you download the possibly long UID list every 30 seconds, and ISTR that the UID comparison code is still O(n^2). > Of course - but they require a specific server setup. I suppose most > if not all ISPs would rather take care that their mail servers can > handle short polling intervals of their customers, than offer an > alternative SMTP solution which adds a lot of complexity for them. > Complexity is more expensive than faster hardware. Probably. > Also, this would only be reasonably easy to configure if the client > machine had a fixed IP address. If it doesn't, ODMR might be an alternative, but again that boils down to polling intervals... > Of course, the ideal solution technically would be a mail server that > delivers mail directly via SMTP as long as this works, and > immediately switches to POP when the recipient's computer isn't > connected to the network anymore. That requires a notion of "being logged in" with authentication that neither SMTP nor POP3 offer. IMAP offers this as long as the server knows the IDLE extension, it's fetchmail's end that falls a bit short here. -- Matthias Andree |