Re: [hybi] Websockets sub-protocol registry

[Julian Reschke <[e-mail]>]

On 2016-03-15 09:52, Takeshi Yoshino wrote:
> On Tue, Mar 15, 2016 at 1:43 AM, Ted Hardie <[e-mail]
> <mailto:[e-mail]>> wrote:
>
>     On Mon, Mar 14, 2016 at 9:18 AM, Julian Reschke
>     <[e-mail] <mailto:[e-mail]>> wrote:
>
>         On 2016-03-14 17:04, Ted Hardie wrote:
>
>             On Mon, Mar 14, 2016 at 8:56 AM, Anne van Kesteren
>             <[e-mail] <mailto:[e-mail]>
>             <mailto:[e-mail] <mailto:[e-mail]>>> wrote:
>
>                  On Mon, Mar 14, 2016 at 4:01 PM, Ted Hardie
>             <[e-mail] <mailto:[e-mail]>
>                  <mailto:[e-mail]
>             <mailto:[e-mail]>>> wrote:
>                  > We simply don't want to assume it; we'd like it to be
>             specified, at least in
>                  > the RTCWEB case.
>
>                  In the typical use of the word, "a" does not match "A".
>
>
>                  > Back to the IANA side of the question: Any objection
>             to specifying that IANA
>                  > should not register two values that differed only in
>             case?  The current
>                  > instructions aren't clear on that.
>
>                  Why? That would restrict the value space. We don't do
>             that for HTTP
>                  verbs either (although it might make sense there, and
>             require them to
>                  be uppercase).
>
>
>             The IANA registry currently has 23 items in it; restricting
>             the value
>             space does not seem like a current issue.  Basically, you
>             would do it so
>             that confusion in which was meant did not occur in the event
>             that
>             someone did attempt the registration of XMPP (rather than
>             xmpp, which is
>             already registered).   This is a first-come-first-served
>             registry, so
>             there is no one to give helpful advice to registrants on
>             this to avoid;
>             the best we can do is give instructions to IANA.
>
>             regards,
>
>             Ted
>
>
>         The main risk I see is that people could mis-read this as
>         meaning that the values are indeed case-insensitive. Of course
>         that could be mitigated by being crystal clear in the wording.
>
>
>     How would this work:
>
>     "The tokens registered in the Websockets sub-protocol registry
>     created by RFC 6445 Section 11.5 are matched using case-sensitive
>     string match.  IANA is, however, instructed to decline registrations
>     in the registry which differ only as to case, in order to minimize
>     potential confusion among different registered versions.  For other
>     useful advice on avoiding collision, registrants are encourage to
>     consult the non-normative section 1.9 of RFC 6445."
>
>
> Sounds good to me.

+1.

> I agree that there's no normative text specifying what comparison method
> to use.
>
>
> FYI, current status of browsers:
>
> The implementation of the WHATWG WebSocket API on Chromium 49.0.2623.87:
> - validates that the Sec-WebSocket-Protocol header value in the
> handshake response is one of the subprotocols listed in the handshake
> request.
> - Case-sensitive comparison is used for this.
> - When the validation fails, the handshake fails.
>
> The API implemented on Firefox 44.0.2:
> - also validates the value against the list sent in the handshake request.
> - But case-insensitive comparison is used.

...raise a bug report?


 > ...

Best regards, Julian