Re: SPDY as HTTP replacement?

On Wed, 7 Dec 2011, Larry Masinter wrote:

> This message had the wrong "subject" line, and didn't note that it was also in response to Yves' ACTION-618....
[..]
> I'd like a discussion of HTTP and SPDY if only to review Yves' summary ...
>
> I remember coming away from the TAG F2F presentation at the F2F with the 
> impression that SPDY was optimized for a relatively narrow set of use 
> cases compared to the full breadth of current HTTP applications, that 
> SPDY would serve well in those cases (a high-performance 'upgrade' 
> option for those servers that matched the use case for which it was 
> designed), but that SPDY was also far from being a HTTP replacement.  I 
> think if SPDY were *only* useful for the "top 100 sites on the 
> internet", it would be still worth developing and bringing to standard.
>
> This wasn't based on a technical analysis of SPDY and counter-examples, 
> but rather that (as it seemed in the Q&A session after the presentation) 
> that the analysis and optimization against the HTTP-NG failure cases 
> hadn't really been done.
>
> To be clear, I think SPDY could succeed and be an important optimization 
> of the Internet, and I'm all for further development, standardization, 
> analysis and further work, I just don't imagine we are close to calling 
> it HTTP 2.0.

The point is not to call it HTTP/2.0 but if it can become a kind of 
"optimized serialization on the wire" of HTTP.

> Now, maybe this is just my personal recollection? Did I miss something 
> or mis-remember? The minutes of the discussion at the F2F are sparse.
>
> Also, since a great deal of the HTTP-NG work was done by W3C before the 
> project was abandoned, perhaps there might be some organizational memory 
> that would help.... a review of "why HTTP-NG failed" would be helpful; 
> my memory is a little fuzzy about it, although at the time, I was 
> disappointed that the project was abandoned.

HTTP-NG ranged from optimized network usage (see webmux that I referred 
to) to uniform modeling multi-layered approach, a bit like features and 
properties in SOAP that also never took off (how to negotiate features 
and decide what to use when is an example of an issue)

Webmux is the closest to what SPDY does here, and I know that SPDY authors 
looked at it (see
http://groups.google.com/group/chromium-discuss/browse_thread/thread/ae93b9f77b833216/684bc7a85c595f9b 
)

-- 
Baroula que barouleras, au tiéu toujou t'entourneras.

         ~~Yves

Received on Thursday, 8 December 2011 14:25:12 UTC