On Fri, 11 Dec 2015, Lucas Pardue wrote:

> I agree with the extra size (although it is all relative) but are there
> really additional round-trips?

Eh, sorry that wasn't very clear of me. It isn't necessarily *extra*
round-trips, but it may require a few round-trips before HTTP/2 is finally
used. Like for example many HTTP/2 servers don't do the upgrade on POST.

> I've been trying to find the time to look at 'h2c prior knowledge' in
> curl/libcurl, still some way to go. Prior knowledge avoids upgrade overhead
> but would need an option to signal usage, I think the new argument creates a
> pathway for such an option.

I think you're right. Once this goes in, I think we can have a look at adding
another option and do prior-knowledge that way.

(I have the code mostly done for 2TLS, I just fell over a test failure I need
to sort out before I land this.)

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html