Hello WG,
(Chris, can you add this as an issue?)
I have found some doubt in myself about how to calc the hashes and
signatures. And by rereading draft-12 I was not able to solve them.
So, I want to share my doubt, and give an (one of several)
solutions. Also, I propose some "new text", to be included in the
(draft) RFC
I give this new text, so we can study it, (given the hints in this
mail), and be sure the sign-RFC will be clear.
Both the solution(s) and the proposed text is "as I think the WG has
meant it to be", with 1 exception, That one (no 3) is marked.
In short, some general questions/solutions:
1) A hash of a (normal) syslog-message is over the complete message,
including field-separating spaces
In [3.9] this isn't specified. While [3.10] says "excluding the
spaces" when mentioning signatures.
2) The signature, both in a SignatureBlock and a CertificateBlock, is
calculated over the hash of the message. Not over message itself
In [3.10] and [4.3.8] this is vage. It say "the signature". In
other code and in (the mentioned_ standards about
signature-algorithms the signature is always calculated over a
hash.
3) The hash, used for signing the SignatureBlock and the
CertificateBlock, is calculated over all field excluding the
signature-field, and excluding spaces direct before the
signature-field. (the field separator). It includes the PRI-part,
the HEADER-part and all MSG-part fields upto and including the
HASHES c.q. fragment-length, including the spaces between them.
NOTE: This is different from the proposed text.
Which excludes all spaces that are fields-separators. That
however is harder to implement, but adds no security or
functionality.
To implement "without sep-spaces", the code has to decided for
each space whether it is a "used a space" of used as field
separator. Doable, but complex.
Also, it requires to algorithms to calculated a hash. One with
"all chars", and one "skipping spaces". As said, needless.
Note: the references above [x.y] are section-number in draft-12
Text to be included:
"""
Calculating Hashes and Signatures
Before a Signature or Certificate Block can be send, some
cryptographic calculations needs to be done. Elsewhere in this
document is specified which algorithms need to be used, and
where to place the result. This section specifies the data used
as input for those calculations
For each device-message (not for relayed messages), a hash SHOULD
be calculated. It is REQUIRED to use the complete message
including PRI, HEADER and MSG parts as input for the hashing.
Those hashes are transited, later, in a Signature Block.
Both, the Signature Block and the Certificate Block contain a
digital signature. Those signatures SHOULD be calculated over the
HASH of the partially composed message. It is REQUIRED to
calculate the HASH of all parts and all fields of the composing
message, but the signature-field. Also, the separating space(s)
direct before the signature-field NOT SHOULD be part of this
calculation. After calculating the HASH and the SIGNATURE, a
space and the SIGNATURE should appemded to the message. It is
RECOMMENDED to send this message directly, as the timestamp will
age.
"""
I think this can be included in chapter 5, as 5.1.*; where the
existing text of chapter 5 become 5.2.* And the name of chapter 5
becomes "details"
--
ALbert Mietus
Send prive mail to: [EMAIL PROTECTED]
Send business mail to: [EMAIL PROTECTED]
Don't send spam mail!
