[comp.protocols.tcp-ip] Unknown TCP/IP options

drw@pascal.mit.edu (Dale R. Worley) (08/31/90)

Is there any standard for how a TCP/IP implementation ignores options
that it does not understand and/or implement?

Of course, that doesn't happen now, because there are only three
options, and all implementations are required to implement them, but
this might not always be true.  What I'm proposing is that an
implementation can ignore an unimplemnted option, which requires that
there be a convention to tell whether an option has an argument or
not.  If I remember correctly, among the current options, 0 and 1
don't have arguments, and 2 does, so we should establish the
convention that the '2' bit (second-lowest) is 0 if the option has no
argument and 1 if it does.

Is this a good idea?

Is it worth worrying about?

(Please reply by e-mail, as I don't read this newsgroup regularly.)

Dale Worley
drw@math.mit.edu

zweig@casca.cs.uiuc.edu (Johnny Zweig) (08/31/90)

Ignoring Unimplemented TCP Options -- it's not just a good idea; it's the
Law.

postel@VENERA.ISI.EDU (08/31/90)

Dale Worley:

The statement has been made that the only options that will ever exist
with out a length field are the end-of-options (0) and no-operation (1)
options.  All other options now and defined in the future will have a
length field as their second byte.

	"					so we should establish the
	convention that the '2' bit (second-lowest) is 0 if the option has no
	argument and 1 if it does.

	Is this a good idea?"

NO.  This is a bad idea.  Just treat special case option codes 0 and 1. Look
for the length field for all other option codes.

--jon.

braden@VENERA.ISI.EDU (08/31/90)

	From tcp-ip-RELAY@NIC.DDN.MIL Fri Aug 31 08:43:17 1990
	Date: 31 Aug 90 05:23:28 GMT
	From: math.mit.edu!drw@bloom-beacon.mit.edu  (Dale R. Worley)
	Organization: MIT Dept. of Mathematics, Cambridge, MA, USA
	Subject: Unknown TCP/IP options
	Sender: tcp-ip-relay@nic.ddn.mil
	To: tcp-ip@nic.ddn.mil

	Is there any standard for how a TCP/IP implementation ignores options
	that it does not understand and/or implement?

Dale,

Yes.  Please see sections 3.2.1.8, 4.1.3.2, and 4.2.2.5 of RFC-1122 for option
processing in IP, UDP, and TCP, respectively.

Bob Braden