- From: Tom Stephens <tomste@microsoft.com>
- Date: Sat, 12 Oct 1996 16:59:46 -0700
- To: "'Eric Murray'" <ericm@lne.com>
- Cc: "'tomw@netscape.com'" <tomw@netscape.com>, "'marcvh@aventail.com'" <marcvh@aventail.com>, "'ietf-tls@w3.org'" <ietf-tls@w3.org>
Eric: Yes, we did review the latest password authentication scheme with the NSA before posting it. And I think there may be some confusion that we are proposing this scheme only to circumvent US crypto export policy. Not true. The real objective is to have a *standard* way to do passwords within the TLS protocol. There's a lot more to it than "non-export crypto" - but just about everything has already been said in this discussion over the past couple of weeks. Tom Stephens tomste@microsoft.com >---------- >From: Eric Murray[SMTP:ericm@lne.com] >Sent: Friday, October 11, 1996 12:44 PM >To: marcvh@aventail.com >Cc: tomw@netscape.com; ietf-tls@w3.org >Subject: Re: Closing on shared-key authentication > >Marc VanHeyningen writes: >> >> > No, you should certainly do something more than just send the password >> > encrypted. You should avoid sending the password at all, encrypted or >> > otherwise. Some sort of challenge/response mechanism would be >> > appropriate, but you are protected from eavesdroppers if you encrypt >> > the data. >> >> True. I'm clearly misunderstanding you then. You said previously: >> >> >There is no need to add a mechanism >> >to TLS when all existing protocols already have a password mechanims. >> >> I assumed the password mechanisms that you meant there were >> cleartext ones, not more sophisticated ones based on challenge-response >> or keyed hashes or anything else. Was I wrong? > >Sort of. They're cleartext unless the entire exchange is protected >by TLS. Then they're encrypted in whatever ciphersuite TLS >negotiated. Obviously when you are negotiating use/non-use >of TLS in an existing protocol, you must start TLS before >sending the username/password. > >> I believe there is a need to add a mechanism to TLS because, while all >> existing protocols have password mechanisms, they're lousy ones. > >The only advantage to embedding passwords is being able to >use "non-export" encryption on the password, _IF_ the protocol >is designed so that no one can use the password field to >exchange "random data". If I can write an app to use >TLS/password auth to send words (say as "login attempts") >of my own choosing under strong encryption, the NSA will not >allow export of the scheme. I have not reviewed the latest >TLS password proposal, so I do not know if it will meet >the NSA's requirements. Has anyone asked them yet? > >In any case, I agree with Tom that we should not be designing >the TLS protocol to meet the US crypto policy flavor-of-the-month. >Besides the fact that they can change it on any political >whim, rendering our work invalid, TLS is an international >standard. > > >-- >Eric Murray ericm@lne.com ericm@motorcycle.com http://www.lne.com/ericm >PGP keyid:E03F65E5 fingerprint:50 B0 A2 4C 7D 86 FC 03 92 E8 AC E6 7E 27 29 >AF > >
Received on Saturday, 12 October 1996 20:00:19 UTC