[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Human I&A, IPsec, and their non-relationship
The motivation for per-user keying based on chosen plaintext seems
unconvincing to me. The amount of data sent per key can be controlled
by the OS, and adjusted to a conservative value based on the algorithm
in use. Wouldn't it be cheaper and safer to rekey host-host
connections than to negotiate and rekey many user/host keys?
BTW, the per-user keying isn't particularly a problem for the xkernel,
as our crypto extensions do key management based on a very general
addressing form, and we've worked with user id's in implementing
Kerberos as a protocol "layer" (albeit a complicated layer). But we
were still able to keep all this out of the transport layer. It
would be "cleaner" for all IPSP implementations if the user id stayed
above the transport layer and the SAID's stayed below.
Follow-Ups:
References: