In message <[email protected]>, Vitor Manuel Basto Fernandes writes:
...
>I am looking for any implementation of RFC 1693 - "An extension to
>TCP: Partial Order Service" (preference for Linux).
>I would like to find out its improvements over "non partial order TCP" on Long
> Fat Pipe Networks.
>The implementation of RFC 1693 does't seem to be a light task. TCP is a byte
>stream protocol and this partial order implementation requires a message
>framework, moreover flow control, error recovery, etc, would require a
>substantial change. I guess backward compatibility would become complicate as
>well on this scenario.
>At the moment I am wondering if its worth for me to try its implementation...
>So if any of you have some hints on this subject please let me know.
>
>Thanks in advance,
>Vitor.
Vitor,
As you point out, making partial order service live within the TCP
framework presents some difficult problems in all the areas you
mentioned. I'm completing my Ph.D. thesis on partial order service,
which explores these very issues you named above. I am doing
performance experiments with an implementation of a partial order
service that is similar to, though not identical to RFC 1693. Right
now it is for Solaris, but I plan a Linux port this Fall, as soon as
the dissertation work is complete. Check back at my web site
http://www.cis.udel.edu/~pconrad in 2-3 months; more complete
information should be available then, including a copy of the
dissertation.
Phill Conrad ([email protected])
Ph.D. Candidate -- Dept. of Computer and Info Science -- University of Delaware
>
>--------------------------
>Vitor Fernandes
>Pos-graduation student
>on computer communications
>University of Minho
>Portugal
>--------------------------
>
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:45 EST