Re: [colorforth] End-to-End Internet Packet Dynamics
- Subject: Re: [colorforth] End-to-End Internet Packet Dynamics
- From: Mark Slicker <maslicke@xxxxxxxxxxx>
- Date: Wed, 28 Apr 2004 01:22:59 -0400 (EDT)
On Tue, 27 Apr 2004, Kurt B. Kaiser wrote:
> Mark Slicker <maslicke@xxxxxxxxxxx> writes:
>
> > I have not yet finished it. Perhaps a TCP implementation could be informed
> > by such a study.
>
> I think this paper has a good discussion of the difficulties and pitfalls
> awaiting a TCP implementation. However, much more is needed.
>
> For an intro, try
>
> M. J. Donahoo and K. L. Calvert, TCP/IP Sockets in C.
>
> That book is 130pp, $14.95
>
> However, in its introduction, the authors say, "This book will not make
> you an expert! That takes years of experience, as well as other, more
> comprehensive, resources." They then cite the Stevens books.
>
> And I recollect a passage in one of the Stevens books which advised
> that trying to implement solid TCP/IP code from a book was not
> possible, that it was necessary to study working code to be
> successful.
>
Thanks Kurt, for the references, I'll have see if I can check these out at
the library.
> > I don't yet understand the meaning or purpose of "window", how often
> >is packet loss due to having no where to put the packet? Does
> >"window" determine the data rate in any other way than as an on/off
> >switch?
>
> www.ssfnet.org/Exchange/tcp/tcpTutorialNotes.html
Yes that confirms my interpretation. The TCP can only send number of bytes
contained in the window until the next ACK.
I question why they didn't use a data rate for flow control. Also all the
ACKs returning to the server seems wasteful, especially in
unidirectional bulk transfers which are the most common case. If the
bytestream was not considered infinite but had check points, the reciever
could selectivly request packets, ACK being implied in the common case.
Mark
---------------------------------------------------------------------
To unsubscribe, e-mail: colorforth-unsubscribe@xxxxxxxxxxxxxxxxxx
For additional commands, e-mail: colorforth-help@xxxxxxxxxxxxxxxxxx
Main web page - http://www.colorforth.com