Re: [core] draft-hartke-core-stateless

[Christian Amsüss <[e-mail]>]

Hello Klaus,

On Tue, Sep 11, 2018 at 01:31:14PM +0000, Klaus Hartke wrote:
> But proxies also generate responses themselves, e.g., from cached
> responses or in case of errors (5.02, 5.04). We don't want clients to
> receive the extended token from another client or no extended token at
> all when a proxy does not support the option.

As long as the option is part of the cache key, any cached response
would need to have a matching request Extended-Token and thus would have
a matching response Extended-Token as well.

If those, by any chance, match up between two different staetless
proxies, *and* if there is a cached fresh response available, then so be
it, may the new request receive the original response. After all, it is
fresh, and the response Extended-Token returned is exactly the one
requested. If both requesting stateless proxies were stateful ones,
neither request would have a Extended-Token option, and the second
client would receive the cached request just as well.

(This is about the astronomically unlikely case of a collision in the
integrity protected tokens; it's way more likely that the stateless
proxy re-requested the resource on behalf of the very same origin if
such a cache hit happens. But still, even if, it'd be OK).

Best regards
Christian

-- 
To use raw power is to make yourself infinitely vulnerable to greater powers.
  -- Bene Gesserit axiom