From: Hallam-Baker <hallam@ai.mit.edu>
Date: Fri, 13 Sep 1996 09:07:44 +0800
To: cypherpunks@toad.com
Subject: Re: [Long] A history of Netscape/MSIE problems
In-Reply-To: <517brn$lu7@life.ai.mit.edu>
Message-ID: <3238962F.1372@ai.mit.edu>
MIME-Version: 1.0
Content-Type: text/plain


Some comments:-

1) Netscape were using BSafe but Kipp unhooked the ergodicity testing
code which is meant to detect inadequate random number generation
methods.

Netscape got full access to Bsafe in return for 1% of Netscape stock.


2) Most serious concern was not the 40bit part. That was simply the US
govt making the security a joke and there was nothing that Kipp or
anyone else could do about it.

More serious were the structural problems, SSL cannot be used across a
firewall (unless the admin is a twit) because it is entirely opaque. SSL
is not very good as an authentication only option.

There is no facility for escrow of keys - another essential feature if
you are to use it inside a corporation. If I am the CIO of IBM I'm not
having the company secrets go out the door via some encrypted stream I
can't read. Similarly banks and nuclear power stations have a legit need
to snoop on their own lans.


3) Microsoft proposed PCT because they wanted to force Netscape to make
SSL an open standard rather than one Netscape could tweak as they liked
and freeze Microsoft out of the picture. Quite how they would imagine
that anyone would want to ever do such a thing...


4) The initial weakness exposed in SSL was that integrity attacks had
not been considered at all. It took a while to explain that this was in
fact a more pressing concern than confidentiality in many applications.
The SSL.v2 integrity was not actually bound to a particular site. Simon
Spero produced a rather nice proxy server which allowed one to reroute
URLs but keep the key intact. (ie connect to foobar.com and get the
netscape home page).

There was a long list of security holes in SSL, PCT plugged a good
number of them and SSL v3 plugged a few. The overall design never gave
me confidence however. Like much Netscape stuff they start with an
over-simple view of the problem spec and then try to solve problems by
adding extra ornaments.


		Phill