Re: Alt-Svc #62: proxies

On 1 Jun 2015, at 12:12 pm, Amos Jeffries <squid3@treenet.co.nz> wrote:
> 
> On 1/06/2015 1:01 p.m., Mark Nottingham wrote:
>> <https://github.com/httpwg/http-extensions/issues/62>
>> 
>> The current suggestion in the issue (made by me) is to add something like this to the spec:
>> 
>> "Clients configured to use a proxy SHOULD NOT use alternative services."
>> 
>> Comments?
> 
> +1.

Thinking about it a bit more, I'd actually modify that proposal slightly:

"A client configured to use a proxy for a given request SHOULD NOT send it to an alternative service, but instead use that proxy."

Make sense?


> The ALTSVC frame is explicitly hop-by-hop already. This helps to bring
> the HTTP/1 header behaviour in line with that.
> 
> A proxy will not be emitting ALTSVC frames. Nor should it be relaying
> Alt-Svc headers, unless its a legacy install treating them as unknown
> and relaying blindly.
> 
> I would hope for the HTTP/1 header to also be explicitly mentioned as
> hop-by-hop as well with a proxy / gateway requirement to SHOULD erase
> where possible on forwarding, or if received from an upstream proxy.

That seems reasonable, but I want to make sure we have agreement on that; there may be scenarios where a client would discover alternative services through a proxy, and then use them later (e.g., when the proxy isn't configured, or on a request that it's not configured for through something like proxy.pac). 

I've opened <https://github.com/httpwg/http-extensions/issues/83> to track.

Cheers,


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

Received on Monday, 8 June 2015 01:06:07 UTC