Re: HTTP/2 GREASE, Results, and Implications

[David Benjamin <[e-mail]>]

On Thu, Oct 31, 2019 at 1:18 PM Willy Tarreau <[e-mail]> wrote:

> On Thu, Oct 31, 2019 at 01:08:53PM -0400, Bence Béky wrote:
> > Thanks, Willy, for pointing out the different sections of RFC7540
> > concerning unknown frame types.  It seems to me that for the past few
> days
> > Chrome (on certain channels) has been sending frames on half-closed
> > (remote) streams.  It might be worth fixing that and re-running the
> > experiment.  I'll circle back with results if we end up doing that.
>
> OK. I have a pending patch ready for testing on the haproxy front which
> addresses this mismatch. The only thing is that chrome beta doesn't seem
> to be available on linux so I constantly have to bother other people for
> testing, which takes time :-/
>

Chrome beta (also dev channel) should be available on Linux. See
https://www.google.com/chrome/beta/?platform=linux. (Or if you've already
gotten the apt repository added, I believe the google-chrome-beta package
should work.) It also uses a separate profile directory, so it won't
conflict with your usual install. Hopefully that'll ease testing.


> Regarding the impacts on our uesrs of anoyher test, I'm not much worried
> of the risk to break a few sites running unfixed versions of haproxy if
> a new Chrome version enables the test again; we've faced some hard to
> diagnose bugs not too long ago and users are trained to disable H2 if
> something really bad happens. Of course it's not desirable but it's not
> as if everything definitely broke. Also our community tends to be quite
> reactive when it comes to applying important fixes. Thus I full support
> restarting the test ASAP :-)
>
> Cheers,
> Willy
>
>