Re: DRAFT - HTTPtre scope of work

to the extent they don't overlap, imo your suggestion is out of scope for
tre.. but it might be in scope for another effort.. we've tried
(unsuccessfully) in the past with a "tuning tcp for h1" document for
example and I think we're open to more work along those lines in other
documents. that's just my opinion - the point of the thread is to get
everyone's opinion on tre, but I think a tight focus like Mark suggests is
imperative.

-P


On Wed, Oct 11, 2017 at 10:12 PM, Willy Tarreau <w@1wt.eu> wrote:

> On Wed, Oct 11, 2017 at 09:14:38PM +0000, Patrick McManus wrote:
> > On Wed, Oct 11, 2017 at 7:48 AM, Willy Tarreau <w@1wt.eu> wrote:
> >
> > >
> > >
> > > On Tue, Oct 10, 2017 at 08:43:10AM -0700, Mark Nottingham wrote:
> > > >
> > > >
> > > > 1)  To clearly separate the version-dependent aspects of HTTP from
> those
> > > that are version-independent, to aid readers and implementers, and
> assist
> > > definition of future protocol versions;
> > > >
> > >
> >
> > [..]
> >
> > >
> > > I think that we should also start to give more recommendations about
> things
> > > that should/should not be done to ease porting.
> > >
> >
> > hopefully these thoughts are complementary!
>
> Yes but they don't completely overlap. Being version agnostic is one thing,
> but optimizing for future versions is another one. This can range from
> trying
> to take more advantage of HPACK for H2 to trying hard to make first
> requests
> fit in a single QUIC packet. This can have some consequences on the design
> choice for future header field names and syntax for instance.
>
> Cheers,
> Willy
>
>