Re: [hybi] Ping/Pong body (was Re: TSV-Directorate review of draft-ietf-hybi-thewebsocketprotocol-07)

[Takeshi Yoshino <[e-mail]>]

Hi,

On Wed, May 25, 2011 at 10:04, Patrick McManus <[e-mail]> wrote:

> I think that's looking at it backwards.
>
> If B cares about having the mandatory pong reply be matched to B's ping
> then B can simply create a non-predictable payload that -07 already
> requires to be echoed in the pong.
>

Hmm, I feel that having 1 byte pre-specified indicator rather simplifies
this and is cheap enough. Yes, preparing something unpredictable solves
this. I think it's fine when user-agents and servers for the service are
under one vendor/people's control. A and B can just negotiate and use
different body. But this is interoperability problem between user-agents and
servers provided by various vendors/people. To be sure that the possibility
of collision between body X which I use for my user-agent and any of server
A, server B, ..., etc. is low enough, maybe I would use random bytes of 3 or
more byte long. It costs than indicator.


> If the implementor cares about such things then those bodies are
> extremely poor implementation decisions. It makes sense to add some
>

Yes.


> guidance to the text to help them out, but there is no reason to add
>

Agreed. At least some advice should be given.


> constraints here or change the existing protocol imo. If the implementor
> changes their mind about being interested in such things they can change
> the implementation of their ping body at the same time.
>

Punting this and decide later is one option, but I'm not sure if it works
with small effort of code/spec change and coordination. Different from
extension etc. which will be defined in the future and then implemented,
Ping/Pong will be implemented by everyone based on the current text.

Thanks,
Takeshi