Re: the Host: saga

Maurizio Codogno wrote:
>There's still something I don't understand in the whole Host: story.
>
>First of all, it seems to me that there is a consensus about having the full
>URL in HTTP/2.0, and Roy frowns upon dictating it in 1.1 because of the
>longer deployment time. But what if people write in the specs that 
>1.1 servers MUST recognize both form, and 1.1 clients SHOULD send full URL,
>but MUST fall back to the Host: form in case of a 404 error sent by a
>1.0 server? True, there is a bandwidth waste, but it should not be
>relevant. And in this way, the transition to 2.0 will be smoother.

I still can't imagine any major browser vendor sending the full URL on the
first request right now. At least not unless John Klensin writes me a $5
million contract to provide him with such a browser. But normal users are
just not going to be picking software which is twice as slow at getting the
initial document display on the current installed base of servers.

I do strongly agree with saying that HTTP 1.1 servers MUST recognize both forms.

>The second thing which puzzles me is the statement by Harald, namely
>that the Host: solution
>
>% - STILL loses the method information.
>
>What's the advantage of having the method information? I cannot see 
>the utility of having a http server which understands other methods - I'd
>preder to leave them to other servers on other ports. Surely I forgot
>something trivial: could someone please explain this to me?

Well one use of sending method information is for proxy-servers. For example
we use a proxy server which receives HTTP requests with FTP URL's and then
gets the file and returns it in an HTTP response. But then again its normal
to send full URLs to proxies anyway.



Alex Hopmann
ResNova Software, Inc.
hopmann@holonet.net

Received on Tuesday, 26 March 1996 18:06:38 UTC