Re: Last Call: 'Internet X.509 Public Key Infrastructure Warranty Certificate Extension' to Informational RFC

[Denis Pinkas <Denis.Pinkas@bull.net>]

> The IESG has received a request from the Public-Key Infrastructure (X.509) 
> WG to consider 'Internet X.509 Public Key Infrastructure Warranty 
> Certificate Extension' <draft-ietf-pkix-warranty-extn-03.txt> as an 
> Informational RFC. 
>
> The IESG plans to make a decision in the next few weeks, and solicits
> final comments on this action.  Please send any comments to the
> iesg@ietf.org or ietf@ietf.org mailing lists by 2003-08-26.


I am quite surprised to see that the IESG received a request from the PKIX 
chairs while a discussion still takes place on the PKIX mailing list on the 
document.

The last response from the lead editor (Alice Sturgeon) is dated August 13. 
She recognizes that at least some "clarifications" are needed.

The e-mail is duplicated below and exhibits that there is good progress but 
still no consensus on the document.

At least some changes to the June version (version 3) are expected before 
the document can be published.

Regards,

Denis

Note: I am back from holidays, just today (i.e. August 26 th).

===============================================================================

Hello Denis,

Please see my comments inserted below, as [AS2], to distinguish from the 
first set of replies to your original comments.

Alice

Alice Sturgeon
Chair, Canadian Advisory Committee - Information Technology Security
ISO/IEC JTC 1/SC 27
and
System Policy Architect & Manager of Standards
SPYRUS

-----Original Message-----
From: owner-ietf-pkix@mail.imc.org
[mailto:owner-ietf-pkix@mail.imc.org]On Behalf Of Denis Pinkas
Sent: Tuesday, July 29, 2003 6:38 AM
To: asturgeon@spyrus.com
Cc: ietf-pkix@imc.org
Subject: Re: I-D ACTION:draft-ietf-pkix-warranty-extn-03.txt



Alice,

Please see my responses below.

Comments on warranty-extn-03.txt (June 2003)

1- On page 2. The text mentions: "A relying party that has a warranty from a
CA".

Does this mean that a relying party must have a contract with the CA to get
advantage of the warranty ?
[Alice] No.

Does this mean that a relying party will get a warranty without a contract
signed with the CA ?
[Alice] Yes.

Does this extension allow to make the difference between the case of a
signed contract and the case of an unsigned contract where the guarantees
could be different?

[Alice] No. If there are any differences in the T&C pertaining to a
contractual arrangement, these differences will be outside of, beyond the
scope, and not pertinent to, the warranty offered in the certificate.  It
applies equally to all relying parties, whether or not a contract has been
signed.

[Denis] Then say something like: "Whether or not relying parties have signed
agreements with CAs, the CA will provide Terms and Conditions for this
warranty which relates to its liability."

[AS2] Alternatively, to make the same point, it could remain implicit.

2 - The difference between the base warranty and the extended warranty is
not crystal clear.

How can a relying party know whether it can have the benefits of the base
warranty or of the extended warranty ?

If the extended warranty is present, then the relying party has the benefit
of the extended warranty.  In short, if it is present, then the relying
party knows that the relying party has the benefit of the extended warranty
by its presence alone.  The syntax shows that the extended warranty MAY be
present:
  Warranty ::= CHOICE  {
        none                 NULL,            -- No warranty provided
        wData                WarrantyData  }  -- Explicit warranty

    WarrantyData ::= SEQUENCE  {
        base                 WarrantyInfo,
        extended             WarrantyInfo OPTIONAL,
        tcURL                TermsAndConditionsURL OPTIONAL  }

If the tcURL is present, a human being might understand the terms and
conditions (if he can understand the local language used), but a computer
program will not be able to do so. This means that computers cannot
understand the difference.

[Alice] The computer does not need to know what is in the T&C.  If the tcURL
is present, it is optionally for the benefit of the human reader. Perhaps
you could suggest some text to clarify this in the I-D if you still think it
is needed.

[Denis] Then say something like: "Warranty extensions are only aimed for
human interpretation and contain data that is inappropriate for computer
based verification schemes, the warranty extension MUST NOT be an active
component in automated certification path validation."

[AS2] But this is not necessarily true. It may be that the RP (i.e., the 
human) has to go to a T&C document if the extended warranty is present, if 
the RP wants to know what exactly is in the T&C, but there should be the 
option of not going to the T&C.  If the extended warranty is not present (as 
noted in the syntax, it is optional), then there is no reason that the 
extension could not be processed by the computer. Therefore, the warranty 
extension is NOT "only aimed for human interpretation...".  This may be 
irrelevant to the point of automated certificate path validation, because 
the extension is non-critical.

3 - There is no security on the information placed at the tcURL parameter
since no hash of the terms and conditions is being used. These conditions
could change overtime and relying parties would not be able to demonstrate
that they have changed.

[Alice]
It seems to me that the time stamp on the certificate would suffice to place
the warranty in a point of time at which the specified terms and conditions
applied; there is no need for the extension to demonstrate that as this is
covered by the certificate itself.  This is no different from the
paper-based world in that if an insurance policy changes its terms, it
cannot do so retroactively to the detriment of clients who have paid for a
specific coverage, unless it notified the clients and compensates them
accordingly.

[Denis] The problem is that the CA could change the policy and the relying
party will be unable to demonstrate that the policy has changed. A
time-stamp token over the certificate would not help. A time-stamp token
over the T&C would help, but the relying party does not necessarily have
access to one TSU. The hash approach is much simpler.

[AS2] A hash function on the T&C?

4 - Is the "Relying party Agreement" a document to signed by the parties or
not ? This should be said.

[Alice]
Use of the term "Relying Party Agreement" is no different from its normal
usage, just as "Certificate Policy" is used in the same context without any
change to the meaning from normal usage, as per RFC 2527 and its successor.
In other words, defining the terms, either Relying Party Agreement or
Certificate Policy is not necessary to the understanding of this extension.

[Denis] The term "Relying Party Agreement" is confusing. Use a wording like
"Terms and Conditions for Relying parties" which does not imply that there
is an agreement signed but that unilaterally the CA provides terms and
conditions. Relying parties may like them or not. If they disagree with
them, then this is still "Terms and Conditions for Relying parties".

[AS2] Maybe so, but the RP has the option of not using or relying on the 
certificate. As I said before, this is normal terminology, and should not be 
confusing; it is not being used with a different meaning.

5 - The WarrantyValidityPeriod is insufficient. Usually there is a period to
claim for a compensation which must be made X months or Y days after the
date of the event which created the damage. It is thus not possible to ask
for a warranty during e.g. the whole validity period of the certificate.
Another time period parameter is missing.

[Alice]
This is exactly what the validity period is stating: that the warranty is
valid for a specified time, which is either the validity period of the
certificate, or another, specified time using the parameters as defined in
the I-D.  It is presented and offered this way, so there is no need for
another validity parameter.  If an offeror wanted to specify a time within
which claims could be made, then a shorter time period than the validity of
the certificate can be used.  This is unlike the paper-based world in the
sense that insurance policies are normally of long duration, whereas this
extension is associated with a certificate that normally has a validity of
two-three years.

[Denis] The semantics of the warranty validity period is not defined ! There
is a need to define more precisely what "the warranty validity period" is.
Here is a proposal along my original text. If it does not fit, please
propose an alternative:

" The warranty validity period is a time period to claim for a compensation
which must be made X months or Y days after the date of the event which
created the damage. It is unrelated to the certificate validity period."

[AS2] What if the warranty validity period is indeed the same as the 
certificate validity period? This was the scenario assumed to be most 
likely, as it is stated in s.2.2, and therefore the option was created to 
allow for a reduction in size of the extension by having NULL in warranty 
validity period in the case where it was the same as the certificate 
validity period.  When it is not the same as the certificate validity 
period, then the parameters available for warranty validity period can be used.

6 - The wType offers only two types of warranty: aggregated and per
transaction.

"Aggregated" means that that once the value of the fulfilled claims reaches
the maximum warranty amount, then no further claim will be fulfilled.

"Per transaction" means that each claim is handled independently but no
individual claim can exceed the maximum warranty amount.

Each type has its own problems:

"Aggregated" is like a race where late claimants will get no compensation at
all because they were not "fast enough". It is questionable whether such a
race condition will be acceptable.
It should be understood that aggregation applies to one warranty and one
claimant.  This is analogous to the paper-based world.  When you are covered
by warranty for X product or service, e.g., health insurance, there is often
an upper limit (viz. the "value") up to which claims will be met for each
claimant; i.e., one insured person.
"Per transaction" means that the CA cannot compute the maximum amount of
money that it may have to give away. Which insurance company will ever
insure a risk for which an upper limit cannot be computed ?

[Alice]
The T&C, as noted in section 1, as covered by insurance, will specify the
maximum. The type of warranty selected depends on the nature of the
transactions, the parties to the transactions (e.g., individuals, large
institutions, etc.).

[Denis] What I am asking for is to allow a combination of both types, which
is currently not possible. See below again.

None of these schemes is acceptable. Some combination should be allowed:

   - a maximum amount per transaction, and
   - a maximum amount per certificate with a maximum time period
     to claim for a compensation after the date of the event
     (no race problem implied).

[AS2] You have a good point. We may need some clarification in s.2.2 to 
explain exactly what the currency amount means.  This is clearly stated (and 
obvious) with respect to aggregate, but not so with per-transaction.  It 
should be indicated that there will be a maximum total of per-transaction 
claims, which would be standard practice, and necessary for the ongoing 
health of the CA.  The total warranty offered for per-transaction claims 
could be expressed as a new parameter indicating number of per-transaction 
claims before the maximum would br reached, which would be simple to 
calculate since the per-transaction maximum is stated; e.g., a sub-line 
after the per-transaction type code, number of transactions as a maximum.

7 - The content of the security consideration should be augmented to
indicate the difference between what a computer program can do and a human
being can do with this extension.

[Alice]
I would have thought this to be blindingly obvious.  We would, however, be
quite willing to consider some proposed words to address this, if you do
think it is necessary.

[Denis] If my text proposal related to comment 2 is inserted, then this is
no more needed.

[AS2] Still willing to consider a proposal, but as you can see from my 
response to your comments on point 2, I don't think this is quite right.

8 - The document is not mature enough and its added value is still
questionable.

[Alice]
We believe that it is mature.  Its value may be questionable to those who do
not want to use it, but given that it is (a) proposed as Informational and
(b) always non-critical, surely its availability for use is innocuous for
those that do not want to use it.  For those who do want to use it, and we
have heard from quite a number who do, it will be useful and valuable.  As
in section 1: "When a certificate contains a warranty certificate extension,
the extension MUST be non-critical, ..."

[Denis]
As you can see, we have progressed, but several issues still need to be
addressed.

[AS2] Yes, we have made progress.

Denis


> File(s) can be obtained via
> http://www.ietf.org/internet-drafts/draft-ietf-pkix-warranty-extn-03.txt