March 27, 2002
Attendees
* Jim Jokl (chair) - Virginia
* Renee Frost - Michigan/Internet2
* Neal McBurnett - Internet2
* Chris Misra - Massachusetts
* Eric Norman - Wisconsin
* Bob Morgan - Washington
* David Wasley - UCOP
* Judith Boettcher - CREN
* Ben Chinowsky (scribe)
- Internet2
Discussion
The minutes of the previous meeting were approved without changes. The group reviewed action items:
* [13-March - Judith will
use Jim's short introduction
to the PKI Lite CP/CPS to
seek review of the document
on the CREN CA list.] -
Done.
* [13-March - John will
send out URLs for more information
on Papyrus.] - Still to
do. There doesn't appear
to be any information on
Papyrus on the Web yet;
Judith sent John's summary
of the project to the list
on March 13.
* [13-March - All will test
the CREN Papyrus CA.] -
In process.
* [13-March - David will
test the CREN Papyrus CA
using a Mac.] - Still to
do.
* [13-March - Jim will add
a row for self-signed cert
capabilities to the S/MIME
clients table.] - Done.
* [13-March - In the S/MIME
requirements document, Jim
will a) add a note explaining
when it's OK to violate
the 13-month maximum cert
lifetime specified in the
PKI Lite cert profile, and
b) in the "Supported
E-Mail Clients" section,
separate Mozilla and Netscape,
remove their version numbers,
and add "(Windows,
Unix, and Macintosh)"
to each.] - Done.
* [27-February - Jim will
work with MACE to find more
reviewers for the PKI Lite
CP/CPS.] - Still to do.
David noted that section 6 of http://www.ietf.org/internet-drafts/draft-ietf-pkix-new-part1-12.txt "punts" on path discovery, though it goes into great detail on what to do with certs once you find them. He suggested that the AuthorityInformationAccess field could contain a pointer to a repository where the CA cert is stored; this could then be used recursively to find the trust anchor.
[AI] David will draft a document describing how to use AuthorityInformationAccess for path discovery; all are encouraged to provide input. David noted that this would not be a quick solution, because it would require all certs to be reissued with the AuthorityInformationAccess field. On the other hand, the code required is relatively easy; if the idea is well received, Cal could become a testbed for it. Bob Morgan noted that the alternative of deploying directories that contain the certs isn't likely to be much easier. The group noted that using AuthorityInformationAccess would remove one of the major reasons for dc= naming, though others -- finding other things besides certs in directories, providing a general replacement for c= o= naming -- would persist. Bob noted that draft-ietf-pkix-new-part1-12.txt is about to become an RFC, so David's document will be a supplement to it, not part of it.
Bob outlined the discussion of the Grid proxy cert proposal at IETF last week. The main idea is that a cert holder delegates powers by generating a key pair and signing a short-term proxy cert, which the delegate uses much like a Kerberos TGT or an X.509 junk cert. This proposal has been written up as an I-D (http://www.ietf.org/internet-drafts/draft-ietf-pkix-proxy-02.txt) and is being promoted as a standard in PKIX. A key point raised in the discussion at IETF is that allowing end-entity certs to sign other certs discards a fundamental principle of X.509; X.509-compliant software can't validate certs created this way. Bob noted that "the phrase 'sacred cow' was used...the phrase 'hard-to-change international standard' was used in return". Three possibilities were explored:
1. 1. setting the CA bit
in end-entity certs (David:
"arrggh") and
applying policy constraints
to those certs to limit
what they can sign.
2. 2. setting up a CA specifically
to issue the short-term
certs, a la KX.509. (Bob
inclines toward this option.)
The Grid folks don't like
the idea of needing another
service for this; they prefer
to use the user's end system.
3. 3. using authZ certs
instead, e.g. with SAML
assertions.
David argued that setting the CA bit is "conceptually the wrong thing to do", as -- currently at least -- it implies that the subject is trustworthy. Bob agreed that this is indeed the central question -- Would setting the CA bit give the end-user too much power? --and noted that current Grid policies probably wouldn't permit it. On the other hand, as David pointed out, allowing non-CAs to sign certs would require path processing software to have a whole different branch of logic to establish what the non-CA signing-cert holder would be allowed to do, and thus what the delegation cert holder could do. Because certs not signed by a CA are not asserting identity, David suggested calling such certs "delegation certs" rather than "proxy certs". The group noted that the absence of a distinction between cert-signing and non-cert-signing keys is the main thing that distinguishes SPKI/SDSI from X.509, so maybe the Grid work points in the direction of using SPKI/SDSI.
There was a short discussion of the possible threat to 1024-bit keys entailed by D.J. Bernstein's recent work on factoring methods; see http://www.inet-one.com/cypherpunks/dir.2002.03.18-2002.03.24/msg00560.html and http://www.inet-one.com/cypherpunks/dir.2002.03.18-2002.03.24/msg00583.html While no attempt was made to reach a formal consensus, TAG's overall level of concern about this issue appears to be quite low. The group agreed that the PKI Lite cert profiles are now ready for last call in TAG, after which they will be circulated for review by a wider group. [AI] Jim will a) verify that the cert profiles are clear on the distinction between campus certs and root certs, and b) announce last call for TAG to make changes to the profiles.
Finally the group discussed
possible applications for
PKI-Lite S/MIME. Judith
suggested survey responses
and elections. Eric noted
that these applications
involve verifying anonymity
rather than identity; David
noted that protection of
anonymity in current non-digital
systems is fairly weak,
offering a low bar for an
electronic substitute to
clear. David is writing
up a electronic voting scenario;
[AI] Bob Morgan will send
David references on electronic
voting, and cc TAG. Neal
pointed the group to an
overview of Internet voting
issues at http://www.ss.ca.gov/executive/ivote/.
Judith suggested homework
submission as an S/MIME
application. David suggested
resume submission, noting
that California already
receives lots of job applications
by email, and that both
signing (for the receiver
to verify authenticity)
and encryption (so the sender's
employer doesn't find out
that they're looking for
another job) would be useful
here. [AI] All will continue
to brainstorm interoperable
end-user applications for
S/MIME.
Action Items
1. [AI] 27-March - David
will draft a document describing
how to use
2. AuthorityInformationAccess
for path discovery; all
are encouraged to provide
input.
3. [AI] 27-March - Jim will
a) verify that the cert
profiles are clear on the
distinction between campus
certs and root certs, and
b) announce last call for
TAG to make changes to the
profiles.
4. [AI] 27-March - Bob Morgan
will send David references
on electronic voting, and
cc TAG.
5. [AI] 27-March - All will
continue to brainstorm interoperable
end-user applications for
S/MIME.
6. [AI] 13-March - John
will send out URLs for more
information on Papyrus.
7. [AI] 13-March - All will
test the CREN Papyrus CA.
8. [AI] 13-March - David
will test the CREN Papyrus
CA using a Mac.
9. [AI] 27-February - Jim
will work with MACE to find
more reviewers for
10. the PKI Lite CP/CPS.
11. [AI] 27-February - Bill
and Ken will pursue getting
Michigan's KCA
12. documentation into NMI
Release 1.0.
13. [AI] 27-February - All
who can help test the KX.509
client on Solaris will
14. contact Bill or Ken.
15. [AI] 13-February - Judith
will check with Michelle
on the status of the
16. Tumbleweed plugin.
17. [AI] 13-February - Jim
will find out what cert
store the SSH.com client
18. uses.
19. [AI] 13-February - Jim
and Deb will draft a letter
to SSH.com, to be signed
by as many representatives
of higher education as possible,
asking that the support
for cert-based authentication
now present in their commercial
version be added to both
the server and the client
in their free version.
20. [AI] 13-February - All
will review the updated
PKI Lite S/MIME requirements
document and send comments
to the list.
21. [AI] 13-February - Updates
to the planned S/MIME clients
table
22. (http://middleware.internet2.edu/hepki-tag/pki-lite/
hepki-tag-pkilite-smime-clients-3.html)
23. Jim will ask Ed if he
will work on Netscape Messenger
column
24. Neal will work on Mozilla,
putting all the information
in one column and noting
any Unix/Windows differences
25. Michelle will look at
Outlook 2000
26. Eric will look at Eudora/Tumbleweed
27. Jim will try to recruit
further contributors to
the table
28. [AI] 16-January - Bob
Morgan and Eric will try
to find out if anyone is
planning to add S/MIME to
pine.
29. [AI] 2-January - Ken
will follow up with the
people responsible for testing
the fix proposed for the
L-Soft signed messages problem.
30. [AI] 19-December - Judith
will draft a scenario for
using S/MIME for homework
submission.
31. [AI] 5-December - Jeff
will have lawyers at MIT
review the legal language
in the draft CPS template.
32. [AI] 5-December - Ed
will find out more about
Dartmouth's timesheet-signing
application, for discussion
on the next call.
33. [AI] 5-December - Keith
will point Wisconsin's deputy
CIO to the posted draft
CPS template.
34. [AI] 10-September -
Eric will a) investigate
and document a problem that
Ed has encountered with
using PKIUser objects to
get certs from LDAP directories
(what the user sees in the
retrieved cert is only a
fingerprint, not cert details),
and b) send the list information
on his experience with cert
retrieval using Internet
Explorer.