Re: [core] CoAP and HTTP - Philosophy and Direction

[Carsten Bormann <[e-mail]>]

One very useful rule that this WG has been following so far for the design of CoAP is: 

	When in doubt, do it like HTTP.

There is absolutely no doubt here, though.  There is nothing we can follow.

On Dec 6, 2010, at 18:07, Angelo P. Castellani wrote:

> HTTP supports chunked Transfer-Encoding,

Yes.

> through this feature is
> possible to push data to clients.

(If you mean "push" in the sense of deferred response:)

No.

Yes, I'm aware of "forever-frames" and all those wonderful asynchronous AJAX hacks.
(I built a little toy based on long-polls in 2005; http://tzi.org:2000/s6.html if you are interested).
These hacks are not a part of HTTP, they are abusing HTTP, and they work only because the idiosyncrasies of certain specific implementations happen to line up just right.

I'll leave the last word to Brian Carpenter, who prefixed his review of draft-loreto-http-bidirectional-05.txt with:

> This is real example of protocol abuse. HTTP wasn't designed for this and doesn't do this properly. 


Gruesse, CArsten