Hi Mark,
Sorry for muddying the waters with my comment regarding XFTP. In my reply, I
made the assumption that Graham was using a (momentary) point-to-point
connection on a private VSAT network. That is the scenario that is used by most
of systems with which I am familiar.
Also, I should have made it clear that I have no real experience with the
technique described in the XFTP paper... just that I was very impressed that you
had found an application-level workaround for an underlying protocol problem. I
can certainly understand how the approach could lead to excessive congestion.
Thank you for the interesting reference to the Floyd and Fall paper.
Best regards,
David Godwin
Petroleum/C-Store Development Engineer
VeriFone, a division of Hewlett-Packard
Office: (727) 953-4072 Fax: (727) 953-4001
http://www.verifone.hp.com
Mailto:[email protected]
> -----Original Message-----
> From: Mark Allman [SMTP:[email protected]]
> Sent: Friday, February 11, 2000 8:23 AM
> To: David Godwin
> Cc: '[email protected]'; [email protected]
> Subject: Re: nt to nt tcpsat problems
>
>
> > Sounds like you might want to try XFTP (see
> > http://roland.grc.nasa.gov/~mallman/papers/wosbis.ps). Also,
> > there are some interesting papers at
> > http://tcpsat.lerc.nasa.gov/tcpsat/papers.html.
>
> Actually, xftp is the *wrong* thing to use. XFTP was an
> experiment. It is inappropriate for use on shared networks, as its
> congestion response is not similar to that used by TCP (it ends up
> being N times more aggressive because you are using N connections).
> See the following paper for a discussion about why using multiple
> flows is not appropriate from a congestion control perspective...
>
>
> Floyd, S., and Fall, K., Promoting the Use of End-to-End
> Congestion Control in the Internet, IEEE/ACM Transactions on
> Networking, August 1999.
> http://www.aciri.org/floyd/end2end-paper.html
>
> allman
>
>
> ---
> http://roland.grc.nasa.gov/~mallman/
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:15:00 EST