Spencer,
Thanks for taking the time to read the draft. Responses are inline.
spt
-----Original Message-----
From: Spencer Dawkins [mailto:spencer(_at_)wonderhamster(_dot_)org]
Sent: Friday, February 29, 2008 12:27 AM
To: General Area Review Team
Cc: Sean Turner; Blake Ramsdell; ietf(_at_)ietf(_dot_)org;
tim(_dot_)polk(_at_)nist(_dot_)gov
Subject: Gen-ART Review of draft-ietf-smime-sha2-03.txt
I have been selected as the General Area Review Team (Gen-ART)
reviewer for this draft (for background on Gen-ART, please see
http://www.alvestrand.no/ietf/gen/art/gen-art-FAQ.html).
Please resolve these comments along with any other Last Call
comments you may receive.
Will do.
Document: draft-ietf-smime-sha2-03.txt
Reviewer: Spencer Dawkins
Review Date: 2008 02 28
IETF LC End Date: 2008-03-07
IESG Telechat date: (if known)
Summary: This document isn't ready for publication as a
Proposed Standard - it's got enough cut-and-paste errors and
apparently-omitted text that I'd think twice about trying to
implement it. And if it has a note that says "normative
reference still in progress, should be brought in line with
normative reference before publishing as an RFC", I'm not sure
why it's being last called now (of course, that decision is
above my pay grade).
Wrt the reference, that draft is being worked in PKIX. Hopefully, it will
progress quickly - I'm hoping for this summer (or sooner) for it to
complete
WG LC and IETF LC. Also, the reference is for object identifiers all of
which were assigned years ago and are stable.
Comments:
Please include any nits listed in
http://www.ietf.org/mail-archive/web/ietf/current/msg50518.html
that I may have missed in this review, by reference :-(
Will do.
When one of the last call comments is "There are obvious
errors (intentionnaly left by the editor in order to know how
many people read the document)", this does not inspire
confidence. I note there is no shepherd writeup in the tracker
yet. The "of for" typo below has been in every version since
00. Who else has read this draft?
This was, I believe, his attempt at humor. See my response to Denis'
email.
Abstract
This document describes the conventions for using the message digest
algorithms SHA-224, SHA-256, SHA-384, SHA-512, as defined in FIPS
Nit - I'm not sure what passes for normal in security drafts,
but I'd expect to see SHA and FIPS expanded on first use in
the abstract, and in the introduction of the document. Ditto
for DSA, RSA, and ECDSA.
I will expand the acronyms.
180-3, with the Cryptographic Message Syntax (CMS). It also
describes
the conventions for using these algorithms with CMS and the
DSA, RSA,
and ECDSA signature algorithms.
Conventions used in this document
Nit - it is odd to see this section as part of the abstract...
For some reason the tool picks up this section as part of the abstract.
It's
got a heading title so I don't think it's in the abstract.
1. Introduction
This document specifies the algorithm identifiers and specifies
parameters for the message digest algorithms SHA-224, SHA-256, SHA-
384, and SHA-512 for use with the Cryptographic Message Syntax (CMS)
[RFC3852]. The message digest algorithms are defined in and [SHS].
Concern: "in and" seems to be missing something.
Denis caught this too. Will fix by removing the "and".
If an implementation chooses to support one of the algorithms
discussed in this document, then the implementation MUST do so as
described in this document.
Concern: this MUST (and the parallel MUST in the next
paragraph) seem odd - do you need to say this?
Hmmm ... I guess not. I'll remove both sentences.
Note that [RFC4231] specifies the conventions for use of for the
Concern: "of for" seems to be missing something.
I will remove "for use of" so the sentence will read: Note that [RFC4231]
specifies the conventions for the message authentication code (MAC)
algorithms ....
message authentication code (MAC) algorithms: HMAC with
SHA-224, HMAC
with SHA-256, HMAC with SHA-384, and HMAC with SHA-512.
2. Message Digest Algorithms
The following addresses the parameters field:
Nit - this sentence isn't clear and isn't required. I'd strike it.
Removed.
There are two possible encodings for the SHA AlgorithmIdentifier
parameters field. The two alternatives arise from the fact
that when
the 1988 syntax for AlgorithmIdentifier was translated into the 1997
syntax, the OPTIONAL associated with the AlgorithmIdentifier
parameters got lost. Later the OPTIONAL was recovered via a defect
report, but by then many people thought that algorithm parameters
were mandatory. Because of this history some implementations encode
parameters as a NULL element and others omit them entirely. The
correct encoding is to omit the parameters field; however,
implementations MUST also handle a SHA AlgorithmIdentifier
parameters
field which contains a NULL.
Nit - I'd describe the normative behavior, and then provide
the history (in a separate paragraph), just so implementers
don't have to scan history looking for normative behavior.
Since this is an update to RFC3370 and we want the same algorithm
parameters
I copied the text directly from there. I'd like to keep it the same so
implementers know that there's nothing new.
Concern - the MUST in this paragraph duplicates clearer
normative text in the next paragraph. I'd remove the last sentence.
See above.
The AlgorithmIdentifier parameters field is OPTIONAL. If present,
the parameters field MUST contain a NULL. Implementations MUST
accept SHA2 AlgorithmIdentifiers with absent parameters.
Implementations MUST accept SHA2 AlgorithmIdentifiers with NULL
parameters. Implementations SHOULD generate SHA2
AlgorithmIdentifiers with absent parameters.
2.4. SHA-512
The SHA-256 message digest algorithm is defined in [SHS]. The
Concern - SHA-512, right?
Denis caught this one too. It will be fixed.
algorithm identifier for SHA-512 is:
id-sha512 OBJECT IDENTIFIER ::= {
joint-iso-itu-t(2) country(16) us(840) organization(1) gov(101)
csor(3) nistalgorithm(4) hashalgs(2) 3 }
The parameters are as specified in Section 2.
3. Signature Algorithms
This section specifies the conventions employed by CMS
implementations that support DSA [DSS], ECDSA [X9.62], and RSA
[RFC2313] with SHA2 algorithms.
Nit - why wait until this late in the document to provide
these references?
The algorithms have been mentioned a bunch of times without
them. Move them to the Introduction?
Will move them to the introduction.
NOTE [to be removed upon publication as an RFC]: NIST has not yet
finalized FIPS 186-3 and there is a chance that the draft may be
changed. This may result in differences between what is documented
in the current version of this document and what is in the FIPS. It
is intended to synchronize the final version of this draft with the
FIPS before publication as an RFC.
Concern: why is this document being Last Called now, if the
expectation is that it's going into REF-WAIT and may change
before it's published?
The WG wanted the note to be safe - The parts of FIPS PUB 186-3 that we
use
has little chance of changing.
3.1. DSA
The algorithm identifier for DSA with SHA-224 signature values is:
Concern: SHA-256, right?
Yes.
id-dsa-with-sha256 OBJECT IDENTIFIER ::= { joint-iso-ccitt(2)
country(16) us(840) organization(1) gov(101) csor(3)
algorithms(4) id-dsa-with-sha2(3) 2 }
When either of these algorithm identifiers is used, the
AlgorithmIdentifier parameters field MUST be absent.
3.3. ECDSA
6. References
6.1. Normative References
[RFC2313] Kaliski, B., "PKCS #1: RSA Encryption Version 1.5", RFC
2313, March 1998.
Concern: ID Nits says "** Obsolete normative reference: RFC
2313 (Obsoleted by RFC 2437)"
I saw this in the nits tools and decided to not update it (note there's
also
RFC3447 that obsoletes 2437). RFC 2437/3447 include a description of
RSAPSS/OAEP that in my opinion (and some others) makes it harder to find
the
v1.5 signature scheme. Since we don't use RSAPSS/OAEP I figured it's all
right to keep the reference to RFC 2313. Since this is an update to RFC
3370
(which was published 4 years after RFC 2437) and RFC 3370 still references
RFC 2313, I also figured it would be okay to not update the reference.
spt