Re: [TLS] Review of draft-santesson-tls-gssapi-00

[Martin Rex <martin.rex at sap.com>]

Jeffrey Altman wrote:
> 
> EKR wrote:
> > Jeffrey Altman <jaltman at secure-endpoints.com> wrote:
> > Sure, but as we're seeing in a number of applications, this doesn't
> > preclude the use of self-signed certs.
>
> No it doesn't.  However, as I've seen in a number of commercial and open
> source products, the authors don't want to go to the trouble of
> writing code to generate certificates as part of the install
> so they install the same self-signed certificate and private key
> on every system in the world.

For an administrative UI, the generation of a new keypair should
always be equal or less work than importing an existing keypair.
The problem with vendors/products installing with a static/fixed
keypair from the installation medium is unfortunately real.

One of the reasons for this is likely shortcoming of the PKI toolkit
(broken APIs) and shortcoming of the PKI toolkit documentation
(it lacks examples to create self-signed credentials for common
usage scenarios (e.g. SSL-Server).


>
> Most users are to naive to know how to replace the certificate,
> probably don't know that they should, and assume that because there
> is a certificate there that they are secure.

One can not blame the users/customers for such a problem. 
Those who ship a pre-generated keypair instead freshly generating
one are to blame.


> 
> Self-signed certificates have their place but they should not be used as
> a bootstrap method for other authentication technologies.

I can not see anything bad about the SSH host key generation
as it is designed one many Unix system today.  If the server is started
without a host key, it simply creates a new keypair on startup.


>
> Too many people make the wrong assumptions just based upon their
> presence.  Please do not insist that they be used as part of a TLS-GSS
> solution.

Please DO!  Up to today TLS interoperability has been great because
pretty much every server had one of the mandatory-to-implement
ciphersuites available for interoperability.

Making it easy or encouraging people to configure future Web-Servers
with TLS_GSSAPI only ciphersuites is equal to dropping
mandatory to implement ciphersuites and traditional strong interoperability
of TLS will be lost forever.


-Martin

_______________________________________________
TLS mailing list
TLS at lists.ietf.org
https://www1.ietf.org/mailman/listinfo/tls