Re: [hybi] RSV bits in extensions

In practise, I believe this means that the RSV bits are of very very
limited use to 3rd party extensions and really have  been reserved for
extensions developed by this WG.    For example RSV1 will effectively
become the compression bit once the per-frame-compession draft is accepted
and RSV1 will only be able to be used by other extensions that are
obviously mutually exclusive to per-frame-compression (such as stream
compression)

I'm not saying this is a bad thing... it's probably good to have these bits
available for standard extensions without the need for the complexity of
dynamic bit allocations.     It just means that 3rd party extension should
really avoid using these bits and put their own flags in the payload....
unless somebody wants to define the unlimited-reserved-bit extension to be
used with other extension :)

cheers

On 21 May 2012 09:46, Takeshi Yoshino <tyoshino@google.com> wrote:

> On Sat, May 12, 2012 at 2:24 AM, Peter Thorson <webmaster@zaphoyd.com>wrote:
>
>> Is the intention of the RSV registration to prevent the reuse of RSV bits
>> or simply to catalog their use and help determine potential conflicts?
>>
>
> My understanding is
> - it's to get the allocation widely known to prevent unexpected
> confliction, complication.
> - it doesn't disallow use of the same bit by multiple extensions
> completely.
>
>
>> In the case that an endpoint gets a handshake request for two extensions
>> with incompatible RSV bit registrations what is the expected behavior? Just
>> pick one? Return an incompatible extension code? Return protocol error?
>> Something else?
>>
>
> I think
> - it's ok to request two or more mutually incompatible extensions
> - it's ok that the server accept the client by picking one of them.
> - if the client requested two or more mutually incompatible extensions and
> the server acked two or more of them, the client must fail.
>
> Takeshi
>
> _______________________________________________
> hybi mailing list
> hybi@ietf.org
> https://www.ietf.org/mailman/listinfo/hybi
>
>