>I think it's better to just change "contains" to "can contain", and similarly
>"includes" to "can include".  The specific rules for different tokens (e.g.,
>local-parts vs. domain names) might be a little different.  For example, base64
>won't work universally on domain names, but base32 would.  So the 
>transformation
>needs to meet certain requirements, and so does the encoding.  But those two
>requirement sets come from different places, so I prefer the partitioning as we
>have it.
>
>As for the length, I think we should be more general and just say there may be
>other constraints that also need to be observed in the replacement.

Sure.  While you're at it there's a typo, date -> data.

>> Second, Section 3 could mention that some transformations use a secret
>> key and thereby recover the definition of "redaction key".  All or some
>> of the Key Management section, as well as the statement on its
>> reasonable length, could then be recovered.
>
>I don't think we want to get into the mechanics of various methods at all.

Please let that sleeping dog lie.

R's,
John
_______________________________________________
marf mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/marf

Reply via email to