Re: #34: Alt-Svc-Used indicator granularity

On 2015-01-21 07:01, Mark Nottingham wrote:
> <https://github.com/httpwg/http-extensions/issues/34>
>
> We took a hum about this in HNL, where it seemed that most people needed more discussion.
>
> The proposals we discussed were:
>
> 1) no alt-svc-used indicator  (some support)
> 2) 1-bit indicator    (no support)
> 3) multi-bit indicator    (a little support - one hummer?)
> 4) full hostname    (some support)
> 5) 1-bit or full hostname   (some support)
> 6) don't know   (most "support")
>
> Interestingly, though, there was no(!) support for the current solution, one bit.
>
> How do people feel now?
>
> My personal take -- From what I can tell, most of the uncertainty here is around the privacy properties of the indicator, and people seem to agree that the risk of abuse is present, but not severe.
>
> That seems to indicate something like "SHOULD send the alternative service hostname in the Alt-Svc-Used header field value, unless the client has been explicitly configured not to send it."
>
> This would allow browsers in privacy mode or similar to not send it, while still giving servers the kind of feedback they need; if a server really needs Alt-Svc-Used, they can 421 or drop the connection, and the client will stay where it was. In that sense, it's sort of in the spirit of #5.
>
> How do people feel about that? Or, are there alternative (hah) proposals?
> ...

Sounds good to me.

Best regards, Julian

Received on Wednesday, 21 January 2015 11:10:38 UTC