With the list discussion having run its course, the WG chairs (David and
Kyle in Mirja's absence) have conferred and reviewed the discussion.
The original question was:
----
- Should tcpinc support TCP simultaneous open? Yes or No
-- If "No," should TCP simultaneous open connection attempts
be abandoned or proceed unencrypted?
----
The WG chairs believe that the tcpinc's WG's rough consensus is that:
- a) there is no requirement for tcpinc to support all possible TCP-SO
cases in all circumstances; but
- b) tcpinc should support TCP-SO when it can do so cheaply; and
- c) tcpinc setup failures for TCP-SO should result in proceeding unencrypted.
Disagreements with this should be surfaced on the list.
Two more observations from the WG chairs:
- 1) The above, particularly "when it can do so cheaply" in b), is
well-aligned with what David Mazieres described as C2:
C2. Require application configuration and use one bit to detect
incompatible configurations and fall back to unencrypted TCP (the
current ENO design, called C2 because it's option #2 in the ENO
draft section 6.2).
We (WG chairs) prefer "Encourage" to "Require" as the first word above.
- 2) Bryan Ford reminds us that attackers will be interested in causing
"fail-open" downgrade-attacks that result in no encryption on connections
that would otherwise have been encrypted, and points out that making
active/passive open look like simultaneous open is a possible attack
vector for doing so. We do not believe this to be a fatal flaw in the
a/b/c rough consensus approach stated above, but this is a concern that
merits careful expert and WG consideration in security analysis of
tcpinc.
Thanks,
--David
----------------------------------------------------
David L. Black, Distinguished Engineer
EMC Corporation, 176 South St., Hopkinton, MA 01748
+1 (508) 293-7953 FAX: +1 (508) 293-7786
[email protected] Mobile: +1 (978) 394-7754
----------------------------------------------------
_______________________________________________
Tcpinc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/tcpinc
