From: Eric Young <eay@mincom.oz.au>
Date: Fri, 25 Aug 95 00:34:10 PDT
To: "Perry E. Metzger" <perry@piermont.com>
Subject: Re: Crypto DLL's/SSLeay 0.4.5
In-Reply-To: <199508241311.JAA13033@frankenstein.piermont.com>
Message-ID: <Pine.SOL.3.91.950825081153.9077E-100000@orb>
MIME-Version: 1.0
Content-Type: text/plain


On Thu, 24 Aug 1995, Perry E. Metzger wrote:
> Eric Young writes:
> > On the PGPphone issue, I Personally I feel SSLphone would be a much 
> > better way of doing things.
> Oh, yeah? No user certificates, no way to verify whats on the other
> end. No assurances that you aren't being tricked into using a weak
> algorithm because negotiation doesn't take place under cover of
> signature. Lots of little potential cracks. Thanks, but no thanks.

:-) Agreed, it depends on how you use SSL and implement it, I have not 
added it yet but I'll put in my library hooks so an application can 
refuse to use certain ciphers that are in the library.  Currently you 
can specify your preference of cipher and there 
is a call to return the cipher being used on an SSL connection.  The most 
recent version of SSLtelnet of ours prints the subject name of the server 
and the cipher being used, just so you can know if you are using RC4-40 :-).

As for authentication, agreed, the key distribution problem for X509 
needs work but still, if the audio is good enough, you should know who is 
on the other end :-).

> This is not to slight your code. I'm slighting the protocol.
none taken, my main support for SSL is that there is minimal work to be 
done to make an application support encryption (+ perhaps authentication) 
over a connection.  This means that any work done to improve the SSL 
library (as in certificate distribution and verification) will instantly 
be able to be added to all applications using that SSL library.  If each 
one of 15 different appliction has a different cipher/authentication 
package, there is 15 times the work to upgrade.
Hell, to put PGP type authentication in SSL would probably not be very hard.
It would require a new certificate type and a new 'verify certificate' 
routine and that would be about it.
Basically I'm a bit lazy, I like to write libraries and then keep on 
reusing them.

> > For phone over modem, authentication is not really required
> And why is that?
Again, if the voice is clean enough, you should know who is at the other end.
If you are talking about a program being at the other end, well thats 
another matter :-).

eric
--
Eric Young                  | Signature removed since it was generating
AARNet: eay@mincom.oz.au    | more followups that the message contents :-)