Re: HTTP/2.0 Magic

Yes. Once things settle down a bit, I'll re-check more final candidates against the full set of servers; I just didn't want to pepper 700,000+ servers with too many requests (because that starts to look like intrusion, which in turn can skew the results as well).

Cheers,


On 23/02/2013, at 10:56 AM, David Morris <dwm@xpasc.com> wrote:

> 
> As I understand this series of tests, you did an initial probe of servers 
> with FOO\r\n
> 
> And continued testing with only those servers which TIMEDout in the first
> test. I think that it might be a risky conclusion that every unrecognized
> probe would receive the same quick CLOSE response. 
> 
> On Thu, 21 Feb 2013, Mark Nottingham wrote:
> 
>> Yep, will try to get to that tomorrow; just wanted to start driving the discussion with data today.
>> 
>> Of course, the source is in github, so you can do that while I sleep, if you like :)
>> 
>> Cheers,
>> 
>> 
>> On 21/02/2013, at 8:57 PM, Willy Tarreau <w@1wt.eu> wrote:
>> 
>>> On Thu, Feb 21, 2013 at 07:46:20PM +1100, Mark Nottingham wrote:
>>>> One more little experiment; the "first" request, followed by something that a settings frame might look like: 
>>>> 
>>>> FOO * HTTP/2.0\r\n\r\n\x80\x00\x00\x04\x80\x00\x00\x00
>>>> 31177 CLOSE
>>>>   298 CONN_ERR
>>>>  3673 TIMEOUT
>>>> 
>>>> Not too bad.
>>> 
>>> Mark, it would be nice to first check how many timeouts we get
>>> from these severs using perfectly valid requests, as I suspect
>>> some of them randomly fail because they might be overloaded.
>>> 
>>> Willy
>>> 
>> 
>> --
>> Mark Nottingham   http://www.mnot.net/
>> 
>> 
>> 
>> 
>> 
> 

--
Mark Nottingham   http://www.mnot.net/

Received on Sunday, 24 February 2013 01:19:26 UTC