Re: [TLS] Security concerns around co-locating TLS and non-secure

[Martin Rex <[e-mail]>]

Michael D'Errico wrote:
> 
> Nicolas Williams wrote:
> > On Tue, Nov 09, 2010 at 01:13:42PM -0800, Bill Frantz wrote:
> >> Michael - Do you have a pointer to the Google research results?
> > 
> > I second that.
> 
> http://www.imperialviolet.org/2010/06/25/overclocking-ssl.html
> 
>      "If there's one point that we want to communicate to
>       the world, it's that SSL/TLS is not computationally
>       expensive any more."

The problem is not that it is computationally expensive, it is the
"TLS tax" imposed by server certs from commercial CAs, their short
lifetime (1year), their non-transferability to other hostnames
(markedroids love to come up with new names every few weeks),
the incompatibility with virtual HTTP hosting (which is significantly
on the rise because of IPv4 addresses shortage) and the additional
help desk impact because of the extremely awful error diagnosis
that common browsers show when encountering network problems
or TLS problems.

SNI doesn't really help, because it is simply not available in ~50%
of the installed base (most of this is Windows XP) and that
number will only gradually decrease in the future.  Most of
the _change_ is from new SNI-capable PCs/notebooks/netbooks 
getting added to the count, much less from the installed 
base getting retired.

The complete absense of reasonable error handling and error
visualization of TLS-enabled software (including web browsers)
to the end users is also a considerable roadblock, substituting
the resolution of TLS-related connectivity issues into random
guessing.  


-Martin