On 4/5/12 6:07 PM, Worley, Dale R (Dale) wrote:
> There's a whole lot of confusion stemming from the fact that
> people frequently *use* the "1#" repetition, but it is defined in
> neither RFC 4234 nor RFC 5234. But it is (as far as I know)
> always taken to mean "a comma-separated list of the specified
> items".
>
> The latest reference I can find without actual work is in RFC 6455:
>
> Sec-WebSocket-Protocol-Client = 1#token
>
> And RFC 6455 normatively references RFC 5234. But it also
> normatively references RFC 2616 for the definition of 1#, and
> RFC 2616 has its own complete ABNF definition (which does define
> '#').
>
> In the context of SIP, 1#thing "should" mean: thing *( COMMA thing )
>
> I suspect that the fact that in different contexts, the "comma" of a
> comma-separated list differ is why there is no standard definition
> of "#" in our ABNF.
>
> Perhaps there should be an id-nit requiring that RFC 5234 ABNF be
> used?
3261 tightened up the use of ABNF and eliminated the use and need for
the # construct. ISTM that everything that extends 3261 syntax should
adhere to the ABNF used in 3261, or a successor to that ABNF that is
backward compatible. IMO it would be appropriate to apply errata to RFCs
that don't do so.
Diversion erred in that it at some point updated its reference from 2543
to 3261, but didn't update its ABNF to match. I don't know if it makes
sense, or is allowed, to file errata against historic RFCs.
Thanks,
Paul
_______________________________________________
Sip-implementors mailing list
[email protected]
https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors
