home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.dcom.cell-relay
- Path: sparky!uunet!mcsun!dxcern!brian
- From: brian@dxcern.cern.ch (Brian Carpenter CERN-CN)
- Subject: Re: Question
- Message-ID: <1992Jul22.130820.11652@dxcern.cern.ch>
- Organization: CERN European Lab for Particle Physics
- References: <1992Jul16.150120.28047@walter.bellcore.com> <21956@venera.isi.edu> <1992Jul17.114053.1@research.ptt.nl> <1992Jul17.160015.7158@dxcern.cern.ch> <1992Jul18.154155.1@research.ptt.nl>
- Date: Wed, 22 Jul 1992 13:08:20 GMT
- Lines: 35
-
- In <1992Jul18.154155.1@research.ptt.nl> walvdrk_r@research.ptt.nl (Kees van der Wal) writes:
-
- >In article <1992Jul17.160015.7158@dxcern.cern.ch>, brian@dxcern.cern.ch (Brian
- >Carpenter CERN-CN) writes:
-
- >>>So the question is "who makes those small 48-56 byte packets? And why is it the
- >>>that size. Is there some kind of "natural" packet size that peaks around those
- >>>values? It's hard for me to believe that.
-
- >> Anyone who histograms data packet sizes finds two peaks, one around
- >> 40-50 bytes and one at some higher value - often 512 bytes, sometimes
- >> 1500 (i.e. full Ethernet packets). Even when running on FDDI which
- >> allows 4Kbyte packets it is very difficult to persuade typical software
- >> to use more than 512 bytes.
-
-
- >I don't argue against the argument that nowadays protocols may produce this
- >size of packets. With "one step back" I meant that we have to re-think what
- >we're doing with the protocols that eventually are supposed to perform some
- >"task". That "task" i.e. the transport of a 1Mbyte digitized picture is then
- >what I see as "the driving force". We'll have to design systems and protocols
- >that support that "task" rather than restricting ourselves for eternity to
- >protocols that don't fit.
-
- >If system designers are desgning ATM equipment that delivers ATM cells with a
- >cell loss probility in the order of 10E-8, it doesn't make sense to me to use a
- >protocol that expects an ACK for every few bytes that is sends.
-
- Kees, when typing or moving my mouse (both on an X terminal), I don't
- know about the protocol, but *I* expect an ACK every few bytes, i.e.
- the X client half a kilometre away is supposed to respond to me. So yes,
- as several people have said, small packets are intrinsic to computer-
- generated traffic.
-
- -- Brian
-