Re: Alt-Svc alternative cache invalidation (ext#16)

On Mon, Aug 18, 2014 at 1:23 PM, Martin Thomson <martin.thomson@gmail.com>
wrote:

>
> I think that we should say that the tuple of (origin, service
> protocol, service endpoint) is the key and that new advertisements
> that match update the expiration time.
>

Would it make more sense for (origin) to be the key with a set of (service
protocol, service endpoint) tuples being the value?  This allows a reset or
correction (or removal?) by publishing a new set of [(service protcol,
service endpoint), ...] values.

On a related note, when sketching out a server-side implementation it
became clear that it is hard to distinguish the separate use-cases
(alternate protocol, port, or endpoint) from the current "Alt-Svc-Used"
value.  This ends up being problematic in cases where a server could
receive Alt-Svc from either direction (from a separate endpoint or just a
protocol+port upgrade on the same endpoint).  One option which would help
here without causing privacy problems would be to make Alt-Svc-Used an
enumeration or bitmask rather than just a "0"/"1" value.

      Erik