The IP over ATM Mailing List Archive by date

[zaumen@hsmpk12a-53.Eng.Sun.COM (Bill Zaumen)]

Curtis wrote

> I'm not trying to bash Sun.  It is just that Sun employees keep
> making claims based on testing with no delay and demonstrating that
> they don't understand TCP dynamics at all.

We understand the dynamics as well as Curtis. The problem is Curtis's
reading comprehension.  We provided the information about what we were
testing, and I thought we were discussing the performance of our ip-atm
implementation coupled with device drivers, atm hardware, signaling
implementation etc. as used by TCP/IP.  Curtis's point has nothing
specific to do with ATM: the effects of round trip delay are
independent of whether ATM or some other technology is used.  Tests
over short distances are important for determining how well our
hardware and software perform, and for that purpose, we needed to get
TCP to send as much per second as possible to make sure that
our drivers and hardware aren't bottlenecks.

As I said in my previous message, another group is responsible for the
TCP implementation itself.  One can contact those at SunSoft for
questions about it.

Bill