Re: Making extensibility cheap

On 4 June 2014 16:30, Tatsuhiro Tsujikawa <tatsuhiro.t@gmail.com> wrote:

> We cannot send WINDOW_UPDATE when transmission is blocked in the middle of
> the other frame's payload due to flow control.


Ah, you are considering sending partial frames?   I don't think that is a
good idea.  If you only have a 4KB window and an 8KB frame, then you are
asking for trouble if you send half the frame and hope a window update will
arrive to let you send the rest of it.

If you don't have sufficient window to send a complete frame or you just
don't send it and that stream is stalled.    It's good that data frames can
be fragmented to smaller frames so that small windows can be used, but it
is not a necessary feature to make flow control work, so long as the window
size will eventually be made larger than max frame size.

That said, having fragmentable extension frames is probably desirable.
Isn't this just a matter of providing an END_SEGMENT flag on the extension
frame for extension that need semantic boundaries can reassemble?  Judging
from Mikes email above this appears to have been previously raised,
discussed and rejected?    But I can't see anything about objecting to
extension fragmentation in the original thread:
http://lists.w3.org/Archives/Public/ietf-http-wg/2013OctDec/0601.html
Anybody got any pointers to any discussion on why this was rejected?

cheers

-- 
Greg Wilkins <gregw@intalio.com>
http://eclipse.org/jetty HTTP, SPDY, Websocket server and client that scales
http://www.webtide.com  advice and support for jetty and cometd.