Re: [xri] trust vs signatures

[Dirk Balfanz <balfanz@google.com>]

I like the idea.

Currently, SimpleSign uses a certuri attribute to point to the cert that can be used to verify the signature. I guess that some profiles would use certificates, while others would not? So perhaps the certuri attribute would have to be replaced by something profile-specific?

Dirk.

On Tue, Jan 6, 2009 at 9:00 AM, Brian Eaton <beaton@google.com> wrote:

Hi folks -

There are two proposals out there on trust and signatures.  I'd like
to outline the similarities and differences and figure out which parts
of each to adopt.

One option is Trust Profiles [1].  The Trust Profiles proposal doesn't
discuss how to sign or verify documents.  It doesn't even discuss who
should sign documents.  Instead it describes a framework for talking
about who should sign documents.  The idea is that different XRD
applications are going to have their own specific needs around trust.
Ideally each application would specify a trust profile that all
implementations of that application would use for establishing trust.

For example, an application dealing exclusively with HTTP authorities
might use an HTTP authority trust profile, while applications dealing
with other authorities and trust schemes (DNS?  DCE?  XRI?  Individual
users?) would define their own trust profiles.  This approach will
hopefully let us achieve both interoperability and flexibility.

The other option is Simple Sign [2].  Simple Sign covers the entire
trust process in one go, discussing both the bits and bytes of signing
and who should sign which documents.  Simple Sign has the advantage of
being simple and concise, but I'm concerned that it lacks the
flexibility to deal with different trust schemes: it assumes that all
applications will use a single approach for deciding who should sign
documents.

I like the Simple Sign approach to signing.  I'm less enthusiastic
about the way Simple Sign talks about who should sign which documents.
 Section 3.2 of the Simple Sign proposal offers one single rule for
signing, but I'm pretty sure that one rule won't work for lots of
applications.  What are those applications going to do for trust?

I'd like to handle this by adopting the signing algorithm from Simple
Sign (sections 1, 2, and 3.1 from the wiki), but replacing section 3.2
of Simple Sign with something more like the Trust Profiles proposal.
Hopefully lots of applications will be able to reuse the signing
scheme, but replace decisions about trust with their own rules as
appropriate.

Cheers,
Brian

[1] Trust Profiles: http://wiki.oasis-open.org/xri/XrdOne/TrustProfiles
[2] Simple Sign: http://wiki.oasis-open.org/xri/XrdOne/SimpleSign

---------------------------------------------------------------------
To unsubscribe from this mail list, you must leave the OASIS TC that
generates this mail.  Follow this link to all your TCs in OASIS at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php