[p2p-hackers] MTU in the real world

[Eugen Leitl]

On Wed, Jun 01, 2005 at 12:03:47PM +0100, Justin Cormack wrote:

> The best (perhaps only sensible) situation for using UDP is where  
> your messages match UDP ones, ie they are small (say 1k payload plus  
> some headers) and it doesnt matter if some are lost (ie you dont even  
> need retry - eg if using error correcting codes or whatever). If you  
> want to use UDP, design your protocol around this (or vice versa).  
> Otherwise you will just be reinventing TCP in UDP.

A good use of UDP would be for fanning out of a search query to a global P2P search
engine (using an amplification cascade) -- it would fit into few 100 Bytes,
needs no ACK, and it could be lossy (nobody expects deterministic results
from a search engine).

Given a query amplification rate of 1:10, and a spidering rate of 1:1000 (each P2P
node is responsible for spidering 1000 others) and a reasonable global graph connectivity
(not too much redundancy) that could be quite usable.

-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a>
______________________________________________________________
ICBM: 48.07100, 11.36820            http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://zgp.org/pipermail/p2p-hackers/attachments/20050601/a63a2c9b/attachment.pgp