RE: Header in TCP Mapping

[Edward Lewis <edlewis@arin.net>]

At 12:00 PM -0400 8/12/02, Hollenbeck, Scott wrote:
>While I appreciate what you're saying, we can't let implementations to I-Ds
>dictate what we can or can't do to get the specifications finished.  Having
>said that, though, I'm inclined to leave things as-is unless someone comes
>up with a real compelling reason for the change.

Looking at IP and TCP, the length fields include the length field 
(noting that the length fields are embedded in the header structure). 
My inclination as a chair is to not change this (other that to clean 
up the byte ordering), i.e., leave the length field to cover itself. 
The reason is that past protocols have done it that way.

'Course, there's a potential counter example in DNS - the TCP version 
of the datagram is preceded by a length field that does not include 
itself.  But in this case, the length field is an addition for the 
connection-oriented version of the protocol only.  The datagram that 
follows is all that appears in the connection-less version, which 
means that the datagram is most likely prepared prior to knowing if 
the length field will be needed.


-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Edward Lewis                                          +1-703-227-9854
ARIN Research Engineer