- From: David Brownell - JavaSoft <david.brownell@Eng.Sun.COM>
- Date: Thu, 5 Dec 1996 11:45:05 -0800
- To: dpkemp@missi.ncsc.mil, ietf-tls@w3.org
> Actually, from a protocol purity point of view, I believe it is
> preferable go implicit (eliminate the ChangeCipherSpec message).
That was where I started, and it's still how I would prefer to go,
in the absence of other complicating issues.
However, I am persuaded by Phil Karlton's note that having this
explicit does facilitate some highly concurrent implementations of
SSL3, where the handshaking and record marking (at least) would be
dealt with by separate processing components. The coordination of
those components is much simplified by this message being explicit.
If the TLS WG wants to support such implementations (e.g. done with
hardware assistance, as in those little black boxes sitting on the
end of dedicated lines), it'd be useful to keep this message in the
protocol ... also, it'd be important to update the protocol spec to
adequately describe the sort of problem which is addressed by this
otherwise superfluous (IMHO) record.
- Dave
p.s. I recognize that other rationales were offered, but those are
not ones that I find I can readily accept.
Received on Thursday, 5 December 1996 14:45:33 UTC