Re: [hybi] Frame size

On Fri, Apr 16, 2010 at 12:43 AM, Justin Erenkrantz
<justin@erenkrantz.com>wrote:

> On Fri, Apr 16, 2010 at 12:22 AM, Thomson, Martin
> <Martin.Thomson@andrew.com> wrote:
> > Responding to specific concerns on framing seems not especially
> productive.
> >
> > Proposal:
> >
> >  - Frame size is indicated up front.
> >  - Frames are binary.
> >  - Frame size be strictly limited (2 octets should suffice).
>

yes, yes, maybe.
If this happened, I'd think that this was a much better protocol  :)
I don't think you waste much by using 32 bits for framesize, but *shrug*

One other thing which exists in SPDY that would make me happy if existed in
this protocol:

version ID (so that any proxy, etc, can ignore drop, etc frames)

-=R

>
> To be clear, this would be 16-bits for the size or a max of 65,535
> bytes in a single frame, correct?
>
> (I'm assuming that you don't mean "2 octets" as being that's all you
> can send in one frame.  *grin*)
>
> >  - Sub-protocol needed if messages are larger than the max frame.
>
> One other note from Ireland is that when httpd and tomcat devs talked
> about small frame sizes, most people thought that frame sizes in this
> range would be artificially limiting - a few folks talked about Comet
> apps they've seen transferring large XML files.  Silly perhaps, but
> that's what the app writers did...trying to get them to change is
> pointless.
>
> IMO, going to 32 bits or even 64 bits simply isn't that much of a
> difference - and may actually be easier if it means removing a
> sub-protocol or extension - the phrase "premature optimization" was
> bandied about in keeping the frame size arbitrarily low.  =)
>
> Anyway, if we flesh out a way up-front to do "jumbo" frames (via an
> extension), I think that'd be fine, but I'd be leery of throwing too
> much into extensions.
>
> Other than that caveat, I'm okay with what you propose.
>
> ...snip...
> > IF used for UTF-8 AND implementer counts characters instead of octets
> THEN framing doesn't work.  (One solution to this problem is to start a
> frame with a known sequence of octets, so that this can be detected.
>  Another "solution" is to not worry about it.)
>
> +1 to "not worry about it".  -- justin
> _______________________________________________
> hybi mailing list
> hybi@ietf.org
> https://www.ietf.org/mailman/listinfo/hybi
>