Re: Fwd: I-D Action: draft-melnikov-rfc2088bis-01.txt
Arnt Gulbrandsen <[email protected]>
| Newsgroups | gmane.ietf.imapext |
|---|---|
| Message-ID | <[email protected]> |
Dave Cridland writes: > Since my figure was plucked from thin air, I'm happy to go > along with anyone else's, but I'm curious as to whether there's > any hard data on message sizes and TCP packet sizes on various > networks. Vast amounts of network data are available if you ask people who work on TCP implementations. I don't, but I keep an eye on it since I have written that a long time ago. The tl;dr: The packet size is 1500 everywhere except on VPNs. 1280 may be considered a lower bound these days. Less than that, and you get in trouble with IPv6, which was okay in 2005 but generates too many customer complaints if you do it in 2015. Sending three packets without asking for ack is considered polite and not harmful (Google did a lot of research on that in order to speed up their search result page), so 3*1500 minus a bit is a safe amount to send. Arnt _______________________________________________ imapext mailing list [email protected] https://www.ietf.org/mailman/listinfo/imapext