Re: Serious design flaw in STARTLS documents

Paul Hoffman / IMC wrote:
> Specifically, if none of the other extensions announced are
> security-related, forcing a new EHLO is wasteful.

Are you sure client implementors are going to correctly determine which
announced extensions are security-critical?  For example, PIPELINING is
potentially security-critical.  If an attacker can convince a client
that a server supports PIPELINING, when in fact the server does not,
this can cause Bad Things to happen.