[ietf-dkim] editorials and nits

[John Levine]

>If the verifier doesn't like the l= tag, they should just reject the
>signature, rather than bother doing the math to verify it.

Don't do that.  In the presumably common case where the message hasn't
had anything added to it in transit, you ignore l= and the signature
verifies just fine.  I suppose as a microoptimization you could
prescan the body and fail if the length is wrong.

>Perhaps we need to more globally describe what we mean by "ignore the
>tag", since paragraph 9 of section 3.2 says, "Unrecognized tags MUST be
>ignored."  In that case, what we want to say is that the verifier MUST
>not take action on the tag, but MUST include it in the hash calculation
>for this DKIM-Signature header field.  Do we need to spell this out?

I hope not.  Anyone who can't figure that out is unlikely to implement
a working verifier no matter what we tell them.

R's,
John