On 2/23/06, James Holderness <[EMAIL PROTECTED]> wrote:
>
> I understand the desire for interoperability, but if you have to make a
> recommendation, why Basic+TLS rather than say Digest without TLS which is
> certainly a lot easier? There may be very convincing security reasons for
> wanting to recommend TLS but I don't know enough about security to know what
> these arguments are. So far the only complaint I've seen against Digest was
> that someone hacking into your password database could use the password
> hashes to spoof a login. IMHO that's not a very convincing argument.

It's not a very convincing argument, but it's not the argument one
would make against Digest. Digest requires storing a password, not a
password hash. That's pretty bad. And all of John's points apply as
well.

When the WG finishes this document, the security ADs are going to look
for 'mandatory-to-implement' security features. That's why we added a
bunch of specifics to the XML Security section in the format document.
Any client in the wild is going to have to implement Basic+TLS, so
make it a requirement for clients. Who cares if some Intranet client
that only does Kerberos isn't compliant? That approach also has the
nice side-effect of ruling out sucky HTTP implementations. For
example, J2ME/MIDP 1.0 clients wouldn't be compliant, because they
don't ship with HTTPS. J2ME/MIDP 2.0 clients ship with HTTPS *and*
sockets, so you can do PUT and DELETE on the socket. It's not too hard
to write, and you can buy a MIDP 2.0 prepaid phone from Boost Mobile
for $50, which includes $10 worth of minutes. That's the cheapest
phone you can get in the US with no credit, so it's basically open to
everyone with regular access to a computer.

--

Robert Sayre

"I would have written a shorter letter, but I did not have the time."

Reply via email to