home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.protocols.tcp-ip:4378 comp.protocols.misc:688 comp.protocols.iso:1120
- Path: sparky!uunet!think.com!barmar
- From: barmar@think.com (Barry Margolin)
- Newsgroups: comp.protocols.tcp-ip,comp.protocols.misc,comp.protocols.iso
- Subject: Re: Why sequence number on bytes instead of packets?
- Date: 10 Sep 1992 21:09:16 GMT
- Organization: Thinking Machines Corporation, Cambridge MA, USA
- Lines: 22
- Message-ID: <18odhsINNhvr@early-bird.think.com>
- References: <STEINAR.HAUG.92Sep10201931@delab.sintef.no>
- NNTP-Posting-Host: telecaster.think.com
-
- In article <STEINAR.HAUG.92Sep10201931@delab.sintef.no> Steinar.Haug@delab.sintef.no (Steinar Haug) writes:
- >I have sort of grown up with TCP/IP, and I feel comfortable with the notion
- >of having sequence numbers on bytes, instead of on packets. I have always
- >assumed this was to make the fragmentation/reassembly process easier.
-
- Fragmentation/reassembly takes place in IP, but sequence numbers are in
- TCP, so they don't have anything to do with each other.
-
- >What are the arguments for sequence numbers on bytes instead of packets,
- >given that the protocol in question will *not* have fragmentation?
-
- When building streams out of datagrams, using byte-oriented sequence
- numbers means that you don't have to retransmit exactly the same packet.
- You can buffer up packets waiting to be retransmitted, and retransmit them
- as a whole rather than in the original separate packets. This reduces the
- number of packets that have to be sent during retransmission, although it
- may increase the total amount of retransmitted data.
- --
- Barry Margolin
- System Manager, Thinking Machines Corp.
-
- barmar@think.com {uunet,harvard}!think!barmar
-