Andre Garzia wrote:

Dar,

Sometimes I wonder why standards exists...

to keep standards writers in a job :-)

doing a simple googling on this topic I grooked that at least *BSD and Windows2000 got settings for TCP Windows for receiving and for sending data, I guess this must just be different buffer sizes ain't so?

Yes - but they're compatible - see my reply to Dar's email (summary - transmit window is an additional constraint over and above the primary config of recv window).

Using the sysctl command line tool one can probe this stuff or one can risk his tcp/ip stack using commands like:

sysctl -w net.inet.tcp.recvspace= Desired value
sysctl -w net.inet.tcp.sendspace=Desired value

this examples are for NetBSD 2.0, and I really don't think one should fiddle with this. One would gain nothing I think, if the client and server got different buffer sizes no one will gain by increasing it's own size right? and the chances of this misbehaving is just too big....

Remember that many senders - in particular, servers capable of high performance - have already had their parameters tweaked, and are limited only by what you can accept - i.e. your receive window.

Probably not an issue at 1Mb or below - but when you get above that - especially in the presence of jitter (variable delay) - you need more than 32k window to allow high-speed but distant servers to get close to filling that pipe to you.

So it's worth setting the receive window in many cases. If you're a server and have memory bandwidth to handle it, it's worth setting the transmit window, in case you are contacted by receivers able to utilize bigger receive windows.

And the chances of this misbehaving are pretty small - TCP is designed to flourish (not merely survive) in the presence of many mismatched configurations...



--
Alex Tweedly       http://www.tweedly.net



--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.322 / Virus Database: 267.3.0 - Release Date: 30/05/2005

_______________________________________________
use-revolution mailing list
[email protected]
http://lists.runrev.com/mailman/listinfo/use-revolution

Reply via email to