[ietf-dkim] Interesting Dupe Signatures

[Murray S. Kucherawy]

> -----Original Message-----
> From: ietf-dkim-bounces at mipassoc.org [mailto:ietf-dkim-
> bounces at mipassoc.org] On Behalf Of hector
> Sent: Sunday, November 01, 2009 7:44 PM
> To: John Levine
> Cc: barryleiba at computer.org; ietf-dkim at mipassoc.org
> Subject: Re: [ietf-dkim] Interesting Dupe Signatures
> 
> But a body hash is calculated per signature because it might have a
> different c14n parsing.  So the ideal optimization would be to reduce
> the body hashing to unique c= c14n types. Given a ridiculous multiple
> signatures case:
> 
>    DKIM-Signature: .... c=relaxed/relaxed;   // dupe
>    DKIM-Signature: .... c=relaxed/relaxed;   // dupe
>    DKIM-Signature: .... c=relaxed/relaxed;   // dupe
>    DKIM-Signature: .... c=simple/relaxed;    // other signer
>    DKIM-Signature: .... c=relaxed/relaxed;   // other signer

You need two header and one body canonicalizations in that example.  And that presumes there's no "l=", which creates another dimension in the matrix of canonicalizations you need to run.

I've had a lot of success doing this sort of common factoring work in my implementation, though I sadly have no data about how much computation work it really saves.