It's interesting to read the very first reactions and evaluations: http://web.archive.org/web/20070423042054/http://cypherpunks...
I even almost managed to find a copy of my old "mystery" mail (mystery because I've been unable to find in the archives, even though I know it has to be in there, as evidenced by the old print-out I have):
http://veps.hypertekst.net/misc/anon-remail/
Did find the thread, though:
http://cypherpunks.venona.com/date/1996/03/msg00168.html
Any chance I could get a copy of the full archive? (Email in profile in case you don't want to make the raw archives available over http).
But then you suggest this could be a reason to throw out the notion of "ciphersuite flexibility".
Aren't these two separate things?
Perhaps the flexibility is good.
Maybe the problem is one of complexity and quality control.
Too many ciphers, and incoporating ones of dubious quality.
I still haven't seen anyone mention the other SSL libraries, e.g., axssl, polarssl, matrixssl, etc.
As for CA "infrastructure", what if the user uses OpenSSL's ca function?
She creates her own CA certificate and key and installs it on her device.
Then she downloads a website's certificate, signs it and installs it on her device.
Regardless of whether a wesbite has a paid-for certificate from a commercial "CA authority", she needs to make the final decision whether or not to trust it.
The user is the ultimate arbiter of which website certificates she wants to sign and install. (Not browser authors.)
Websites just need a central repository to publish their certificates.
They already do this for their "domain names" by having them published in a publicly accessible zone file (ideally, the user can download the zone file, as well as query it piecemeal over a network).
We as users trust that these zone files are accurate: specifically, we assume the IP addresses for the website's nameservers are correct.
The fact that you have to upgrade to TLS 1.2, which includes more than just new ciphersuites, somewhat counterfeits the idea that the ciphersuite mechanism provided much protection.
Ultimately, the protocol might have been just as well off by defining a single ciphersuite and accepting that a break in that ciphersuite would necessitate a protocol update.
I want to cover the cipher suite flexibility quickly, just to point out how much of a fantastic idea it is. The server and client tell each other what their most-prefered ciphers are, then they agree on the strongest one that they both support.
This has massive benefits, for instance, when BEAST et. al. came along and effectively broke block cipher based OpenSSL for a while, we could jump to RC4 for a little while. When that was broke, we all hauled ass back to AES. This gives incredibly flexibility when dealing with many attacks, by simply being able to circumvent the affected parts.
This is frustrating.
All three of the major ciphersuite problems in SSL/TLS (BEAST, RC4, Lucky13) were well-known in the literature prior to their discovery.
The BEAST vulnerability --- chaining IVs across messages instead of making them explicit and thus unpredictable --- was described years before BEAST was published, and was presaged by Phil Rogaway in discussions about IPSEC.
It is possible that RC4 was known to be vulnerable all the way from the moment it was leaked to Usenet. Certainly, the statistical biases were well-known. Not just apocryphally, but in the academic literature.
Lucky13 was a product of MAC-then-encrypt constructions, and an instance of a specific attack Vaudenay described in the early 2000s.
TLS has "ciphersuite flexibility" that has enabled the protocol to hop from one vulnerability to another. Meanwhile, fundamental fixes (either explicit encrypt-then-MAC, or AEAD modes like GCM, or secure native stream ciphers) have demanded fundamental protocol changes, not just new entries in a list of ciphersuites.