[comp.protocols.tcp-ip] TCP extensions

fab@saturn.ACC.COM (Fred Bohle acc_gnsc) (02/09/90)

I am trying to find out who uses TCP options for various extensions
of the protocol.  For example, what implementations use RFC 1072, Extensions
for Long-Delay Paths?  This includes window scaling, selective acknowledgements
and TCP Echo options.  I would like to know any other extensions to the
protocol that various implementations use.  What other RFC's are implemented?
What other extensions are implemented which are not yet RFC's?

This information will be used in considering which options/extensions
will be added to ACC's ACCES/MVS product, so your replies will be used
to improve interoperability.

Fred

------------------------------------------------------------------------
Fred Bohle			EMAIL: fab@saturn.acc.com
ACC				AT&T : 301-290-8100 
10220 Old Columbia Road
Columbia, MD 21046
------------------------------------------------------------------------

craig@NNSC.NSF.NET (Craig Partridge) (02/12/90)

Fred:

    At the Internet Engineering Task Force Meeting last week a working
group made some progress on deciding what to do about the RFC 1072 options.
Here's a quick summary:

    (1) The RFC 1072 window shift option will be used to expand the
    window size (everyone likes it, and implementation is a snap).

    (2) We'll create a new Urgent Pointer option so it can point
    to urgent data anywhere in the expanded window.  (To use it,
    the URG bit is left off -- when the receiver processes options,
    it will see that there is urgent data).

    (3) The WG was divided among two options for expanded the sequence
    and ack space.  Thankfully some folks with supercomputers offerred
    to implement both options and report back this spring.  The two
    options were:

	(a) since we need an URG option anyway, steal the urgent
	field in the TCP header to get 8 more bits for each of the
	sequence # and the ack #.  These would be high order bits.

	(b) don't futz with the header (note that in most cases urgent
	data is likely to be within 2**16 of the current sequence number
	and thus the urgent field is still useful).  Instead, put an
	option at the end of each segment header, which contains an
	additional high order 16-bits of the sequence number and
	the ack number.

Based on experience, the WG will decide which path to take.
    
Craig