Re: User defined SETTINGS frame extensions

On 7 February 2014 13:58, William Chan (陈智昌) <willchan@chromium.org> wrote:
> Unhandled SETTINGS MUST be ignored, though.

I'll note that there was subsequent discussion on this point that
concluded with what we have in the spec.  I know Roberto in particular
was sad at this conclusion, but I got the sense that this was another
case of "done > perfect".

I got the sense that there was a reasonably firm commitment to
preventing extension for frame types.  There might be frame types that
don't interact with the protocol state in any way, but most of the
functions those can, to a large extent at least, be fulfilled at
higher layers.

Forbidding setting extension is less popular, since there is arguably
less risk that settings could be used to break things.  At the same
time, it's hard to imagine something that couldn't be fulfilled
through other avenues.  I feel obliged to point out that we've started
to set the bar pretty high for adding new features, and this could be
conceived of as a new feature.  It would have to come with some fairly
strong justification.

Received on Friday, 7 February 2014 23:24:24 UTC