[p2p-hackers] Optimal UDP Message size for data streaming...

[Matthew Kaufman]

> Basically, given that you've got to make your application 
> work well on small-MTU paths, is it worth the effort to find 
> and use big MTUs or just more practical to find the lowest 
> common denominator?

I know too much about how the Internet is designed (see my bio) to believe
that trying to detect >~1500-byte MTU is a useful exercise to engage in, at
least for several more years, unless you are designing a demonstration for a
very specific closed environment (and your application cares about total
throughput, rather than, say, head-of-line blocking issues on slow links)

And, as has been pointed out, overestimating is worse than underestimating.
My personal choice is 1424 for the data portion of a UDP packet...
(802.3 with LLC (1492 bytes instead of 1500) minus a GRE tunnel overhead (24
bytes, one of the worst tunnel overheads there is) minus a standard IP
header with no options minus a UDP header, rounded down to the block size of
my default symmetric crypto algorithm)

Matthew Kaufman
matthew at matthew.at
http://www.amicima.com