Hi;( h2c )
I'm testing my implementation of using Curl with http2 in direct clear TCP
without going through HTTP/1.1 Upgrade,
I observe in wireshark 3 different packets for Magic, Settings and Headers.
Those can easily be sent in one TCP packet to save the roundtrips.
Testing the same with h2load
On 3/12/2016 6:34 PM, Steve Holme wrote:
On Thu, 10 Mar 2016, Henri Hein wrote:
> I forked the curl sources and implemented CURLOPT_SSL_CTX_FUNCTION for
> the SSPI/Schannel build. Would there be any interest in a pull
> request for this? It's just a few lines of code that I took from the
> cyass
On Thu, 10 Mar 2016, Henri Hein wrote:
> I forked the curl sources and implemented CURLOPT_SSL_CTX_FUNCTION for
> the SSPI/Schannel build. Would there be any interest in a pull
> request for this? It's just a few lines of code that I took from the
> cyassl source. For my purposes, I needed acce
On Fri, 11 Mar 2016, Catlow, Martyn wrote:
The curl api work fine when the function is statically linked and called,
but when the function is loaded and called some of the curl calls appear to
work, but curl_easy_perform() returns
"curl_easy_perform() failed: Out of memory"
Is there any way
On Sat, 12 Mar 2016, Molina wrote:
I think I have some good news!
Excellent! Thanks for continuing your work on getting to the bottom of this.
If you want to take a look at the changes please access here:
https://github.com/Moliholy/curl/commit/c72cddb24216a43516bc10253a4a66566db7088e
That
I think I have some good news!
As I said before, I’ve been working with the window sizes, trying to emulate
what I did with nghttp2 to get the numbers I showed… and it seems to work!
Before the changes it reached ~300KB/s, and after the changes ~1MB/s, which is
the maximum bandwidth I have at ho