My apologies if we've covered this already; I think I may have mentioned
it in a larger email and it got glossed over.

In at least two documents we have phrases like "a compliant v1
implementation MUST NOT expect this [version field] to be 0 (i.e., v1)."

The word "expect" seems weird to me, somehow -- in a universe where v1 is
the only thin widely deployed, we should "expect" that responses will use
that version, in that it is the most likely occurrence.  What we are
trying to prevent is the case where implementations assume that the
version field (or really, the version of the message they are processing)
matches the version of the implementation without validating that
assumption.  Do we want to say instead something like "compliant v1
implementations MUST validate the version field while processing messages
and handle inputs from versions other than v1" or "a compliant v1
implementation MUST NOT assume this field is 0 (i.e., v1) without
validating the version field"?

Is this too trivial an issue to worry about?

-Ben

_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans

Reply via email to