Re: Do we kill the "Host:" header in HTTP/2 ?

On Wed, Jan 30, 2013 at 11:36:45PM +0000, Poul-Henning Kamp wrote:
> Content-Type: text/plain; charset=ISO-8859-1
> --------
> In message <CAOdDvNrrDhyayThYJz+94C+U+K18gtvEGWqwZ-8-+wP6kH5QRw@mail.gmail.com>, Patrick McManus writes:
> 
> >and we can apply a requirement to order things :scheme, :host, :uri in
> >order to facilitate streaming ..
> 
> And then we're back to my underlying question:
> 
> Isn't that the sort of thing we should settle before people spend a lot of
> time on simulating compression and encoding schemes, as this will clearly
> have significant impact on the results of those experiments ?

Both are done more or less in parallel, we should not have a religion
about this. If someone comes up with a smart new way to very efficiently
transport such information that is different from what was agreed the
day before, well, let's study the solution before saying "nah, we agreed
on a different transport method".

I know that my mind is mixed on the subject. As an intermediary implementer
I need to find Host very quickly, and having it with the scheme and URI
would be very convenient. But at the same time, I noticed that it might
or might not be the best solution for compression. Lately when I came up
with an idea of a "base" url, it was something different again and I had
no trouble with that. So let's declare *preference* and find what is best
in the end, then tune it to best fit declared preferences if still
applicable.

Regards,
Willy

Received on Thursday, 31 January 2013 00:19:17 UTC