Re: [hybi] I-D ACTION:draft-ietf-hybi-thewebsocketprotocol-08.txt

[Dirkjan Ochtman <[e-mail]>]

On Thu, Jun 9, 2011 at 09:38, "Andy Green (林安廸)" <[e-mail]> wrote:
> Is there anything or we just crank the default version number up to 8?

It looks like there are at least some differences:

- The extended payload length may now use 64 bits (instead of just 63).
- Invalid client handshake results in HTTP error instead of connection abortion.
- Two new closure status codes (1005 and 1006).
- Specified handling of invalid UTF-8 on the client side.

I note that the remarks from my email (from May 30, [1]) seem to not
have had any impact on this draft. Did my message come to late to make
any changes to this draft, did my suggestions not make sense for
inclusion, or did I perhaps not follow the correct protocol for
suggesting improvements? (I noticed that the required server headers
*are* listed in the section on client acceptance of the server's
opening handshake, but not in the section on the server sending that
handshake.)

Two further questions on things that I feel could be clarified, in draft-08:

- This may be silly, but in the introduction text for the figure
(which is very useful!) in section 4.2, I wonder if it would be
helpful to state explicitly whether the bits are listed in big- or
little-endian order. For bytes, this is defined explicitly in the
descriptions, but I know that I was at least somewhat confused about
this (and got it wrong in my initial implementation). This might just
be my inexperience with diagrams/specs like this, so feel free to
ignore this.

- Section 4.5.1 (about status codes for Close frames) does not mention
anything about masking. I would presume that the status code bytes,
being part of the application data part of the frame, should be
masked, but perhaps it's worth pointing this out explicitly.

Cheers,

Dirkjan

[1] http://www.ietf.org/mail-archive/web/hybi/current/msg07461.html