Re: [hybi] thewebsocketprotocol #28 (new): Fragmentation

On 23 November 2010 06:09, John Tamplin <jat@google.com> wrote:
> This question has to be answered in order to interpret a stream of frames
> properly.  One of the following must be true to have a single unique
> interpretation of a sequence of frames:

> - control frames can't be fragmented

+1

> - control frames cannot be inserted in the middle of a fragmented message

I think this is reasonable to allow in the base protocol.  It would
probably simplify the implementations of things like keep alive pings,
so that these can be sent if there was a long pause between fragements
of a message.    It would also allow a close control frame to be sent
without waiting for a message to complete.

> - there must be a way to differentiate a control fragment from a data
> fragment

We have that already- yes?

> - the receiver must keep an arbitrarily deep stack of in-progress messages

I think this kind of complexity need only be supported by stacks that
support a MUX extension.  Once such an extension is accepted, then I
would expect that control frames and messages fragments from different
channels/streams will be able to be interleaved and well defined rules
will need to  be established for interpreting how control frames
relate to partially sent/received messages.

cheers

> John A. Tamplin
> Software Engineer (GWT), Google
>
> _______________________________________________
> hybi mailing list
> hybi@ietf.org
> https://www.ietf.org/mailman/listinfo/hybi
>
>