Re: public key (was Re: consensus probe.)

["Joseph Galbraith" <galb-list%vandyke.com@localhost>]

"Markus Friedl" <markus%openbsd.org@localhost>
> On Tue, Feb 05, 2002 at 02:39:56PM -0500, Bill Sommerfeld wrote:
> > > Perhaps "(as specified by
> > > public key/cert format)" would be more
> > > appropriate?
> >
> > *if we do that*, we may also need a statement that unless otherwise
> > specified the sig format id == pk/cert format id.
>
> i think this is the only way, since during the initial KEX
> both parties only agree on _one_ single format id for
> the hostkey.

It is possible that the format id for the
public key (x509v3) specifies a signature
format that isn't necessarily of the same
name be used.

For example, we could specify (unambiguously)
that x509v3-sign-rsa keys would use ssh-rsa
signatures.  There is no ambiguity here,
hence my proposal that we spec this flexibly
enough to handle this.

On the other hand, I do agree that we also
need language that says that unless otherwise
specified sig format id is the same as key
format id.

Perhaps the following text would be more
agreeable?

   Signatures are encoded as follows:

     string   signature format identifier (as specified by the public
key/cert format)
     byte[n]  signature blob in format specific encoding

  Public key / certificate formats that do not
  explicitly specify a signature format identifier
  MUST be considered to use the public key / certificate
  format identifier as the signature identifier.

- Joseph