Re: [hybi] Review of draft-ietf-hybi-thewebsocketprotocol-13

[Tobias Oberstein <[e-mail]>]

> Using the above definition, I guess validating the overall message makes
> sense.  The one remaining concern that I would have would be if fragmented
> frames are used for streaming.  An application could start processing data
> and get deep into the middle of a stream before it recognized an error.
> Whereas if each frame had to be UTF-8 valid, then intermediate results
> would all be UTF-8 valid as well.

A _single_ frame can be up to 2^63 octets, and a message can consist of
an infinite sequence of frames.

Requiring frame boundaries to observe utf-8 code point boundaries is
not a solution for "fail fast on invalid utf-8".

An incremental utf-8 validator is one - and works regardless of boundaries
or frame length.