On Fri, May 1, 2015 at 1:25 AM, Richard Barnes <rbar...@mozilla.com> wrote:
> 3. HTTP caching is an important feature for constrained networks.

I think it important to emphasize that the affected case is shared
caching in the form of forward proxies. https doesn't prevent caching
in the browser or caching on site-chosen caching nodes (CDNs). (I know
you know this; this paragraph is for the mailing list.)

Whether shared caching in forward proxies is indeed an important
feature hasn't been properly shown in this thread.

To bring a data point to the thread, data from the network of the
University of Edinburgh (http://www.ltg.ed.ac.uk/~ht/HST_noREST.pdf ;
skip forward to PDF page 14) indicates that even without the action
proposed in this thread to deprecate insecure HTTP, the hit rate in
the university's shared cache is already rather low and getting lower.
Obviously, university networks in Europe don't count as constrained,
but this is likely a best-case scenario of cache hits, since this is a
network whose users one might imagine to have more of a common set of
interests in their use of the network (due to being part of the same
organization) than users who have no organizational commonality and
only have locational commonality.

I think without empirical evidence showing the *current* (as opposed
to arguments from 20 years ago) importance of shared caching on the
supposed "constrained networks"--i.e. empirical evidence showing that
the shared cache hit rate is is a make-or-break deal for actual
present-day networks where the bottleneck is between the ISP [the
location of the shared cache] and the backbone and the bottleneck
can't be fixed e.g. by lighting up more fiber--it doesn't make sense
to put effort into building complications that seek to preserve shared
caching in the encrypted future.

> 5. It may be productive to take some interim steps, such as placing
> limitations on cookies stored by non-HTTPS sites.

Forgetting insecure cookies when quitting Firefox is now
https://bugzilla.mozilla.org/show_bug.cgi?id=1160368

-- 
Henri Sivonen
hsivo...@hsivonen.fi
https://hsivonen.fi/
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform

Reply via email to