> >  However, the same is true for HSTS, CSP, and any other policy-conveying
> >  header.
>
> It's not exactly the same - HSTS does not interpret an absent response
> header as max-age=0, so the headers doesn't need to be sent on every
> response.  HSTS headers are also smaller.


This is a big issue I wouldn't overlook. In my experience looking at web
crawls and trying to determine which sites are consistently setting HSTS,
almost nobody consistently sets the header for every path and subdomain
within a domain. For HSTS, this is okay as pointed out, but for HPKP I
expect this will lead to a lot of accidentally clearing of pins. So we
should think about what makes life easiest for server admins deploying HPKP
in addition to efficiency.

This could be argued either way-if servers aren't disciplined enough to
always set the header, they may not be disciplined enough to always satisfy
their own pins and they'll hang themselves that way. Serving everything
satisfying a set of pins is strictly harder than serving everything over
HSTS, and that's an issue. On the other hand servers will be more motivated
to fix anything they serve not satisfying pins, since clients will lose
access and complain, whereas accidentally clearing client pins by not
setting the header is a siient failure.

Personally I'd advocate strongly for a well-known URI. This is a tragedy of
the commons-if every working group uses a header since this is a little
less work, we end up with lots of headers being set and it being much
harder for everybody to manage than a well-known URI specifying all TLS
policy would be, as well as far less extensible.

As an example, it would be useful during the rollout of Certificate
Transparency if sites can specify that they are a "CT required" domain
before this policy could be turned on by default for every domain in the
world. Should this be another new header "Strict-Certificate-Transparency"?
An extra directive for HSTS? Having a simple JSON file at a well-known URI
to add a new field to makes much more sense.
_______________________________________________
websec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/websec

Reply via email to