If you roll your own protocol to reliably break up a huge packet and reassemble 
it on the other end, your really just reimplementing one of the main features 
of TCP. If you are worried about performance, realize that any performance gain 
from using UDP will probably be negated by implementing a reliable protocol in 
(relatively) slow java code. In other words: TCP already does this, and 
probably does it faster than any custom code you could write in java.
   The exception of course is if you don't mind losing some of the data, in 
which case you don't need an expensive two-way protocol to manage dropped 
packets.

Will 

----- Original Message -----
From: "Ashish" <[email protected]>
To: [email protected]
Sent: Thursday, July 16, 2009 9:01:03 AM GMT -06:00 US/Canada Central
Subject: Re: Help With NioDatagramAcceptor : [The message is larger than the  
maximum supported by the underlying transport]

> split huge data pack into multi pack is your own logic!!!

Well the question has been answered. UDP has a maximum limit beyond
which it can't carry the data.
For your situation, you have to have to define your protocol over UDP.
Basically, you have to fragment the data at client and reassemble at
server. The other alternative would be to use TCP instead.

-- 
thanks
ashish

Blog: http://www.ashishpaliwal.com/blog
My Photo Galleries: http://www.pbase.com/ashishpaliwal

Reply via email to