> If the security people are going to dictate in every detail what we
> can or cannot do, then I think it is their responsibility to make sure
> that they participate in our WGs to give advice us advice in the
> context of our needs and discussions, rather than come down and
> criticize what we have done every time we think we have found an
> acceptable solution.
I would say that the foot is definitely in the other shoe:
If you see that protocol development is being done which will
impact your product, it is definitely in your interest to make
sure that the people doing that development know your requirements.
It is 100% obvious to you that when you consider using TLS, the
TLS group is an interesting forum for you.
It is not at all obvious to the TLS group that there is anything
going on in Internet printing that impacts requirements for TLS.
That group too is an open forum; the reason you are not a member is
because you have chosen not to be.
Most of the people working on TLS are there because they need TLS in
*other* projects; their own requirements are obvious to them, yours
And remember: this is not something they do for a hobby; anything
we ask them to do above what they already do usually comes straight
out of their spare time.
The security requirements are really, really simple:
- THINK about what security your application needs, and provide that
- No plaintext passwords
The requirements that apply to ALL protocols, in ALL aspects are:
- Enough requirements to ensure interoperability
- Avoid use of IPR-tainted technology whenever reasonable.
The last two are NOT security requirements, it's just that the
security area, thanks to RSA, has one of the nastiest examples of this
being a problem.
BTW, I consulted with the security AD about ciphersuites in TLS; he
was quite surprised that anyone would consider using TLS without
security, since this adds quite a bit of overhead over raw TCP.
Ciphersuites are "numbered items", but the intent of the TLS group
is that it should be simple to add new ones; the reason for using
suites rather than feature combinations is that some combinations
of features have non-obvious but serious security flaws.
If you want a ciphersuite with different properties than the existing
set, take this up with the TLS group. ietf-tls at consensus.com; use the
-request convention to subscribe.