> >There are a number of potential cases here:
> >
> >a. Legacy (v0) client speaking to v1 server.
> >b. Legacy (v0) client speaking to v0 server.
> >c. V1 client speaking to v0 server.
> >d. V1 client speaking to v1 server.
> >
> >
> >c. Do we understand how existing v0 client implementations will
> >behave if they receive a v1 packet from a server?
>
> today - no, this is not completely predictable.
>
> Part of 3825bis will add text to give guidance on the desired
> behavior for each scenario.
By "unpredictable" do you mean "untested" or "could vary
between implementations"?
In the case of .11 implementations, the code for the v0
client could be present in firmware or a kernel mode
driver. That implies that mis-behavior can have serious
consequences.
We can address cases c and d in RFC 3825bis, and case b
is inherited from RFC 3825.
Do we have the same freedom with respect to case a?
