[ietf-dkim] t=y is ugly and critical tagsb

[Amir Herzberg]

Small complaint: in draft-ietf-dkim-base-01.txt, the record (section 
3.6.1) defines a `flags` tag identified by `t=`, with the only defined 
value being `y`, standing for `testing DKIM, do not discard message even 
if signature verification fails`.

I guess the original role for `t=y` was for `testing flag`, then 
extended so that the same parameter can be used for future-defined 
flags. Probably a good idea, but leaving the choices `t=y` is ugly (not 
mnemonic).

I know implementors will hate to change but I still think the spec 
should adopt the menmonic `f=t` (for `flags=test`) instead... 
Implementations can also support the old `t=y` tag to avoid glitches. 
This should not cause a problem as the spec says to ignore any undefined 
tags (in both key record and signature).

I'm actually concerned also about this. Suppose you need a new tag to 
the key record or to the signature, and this tag is `critical` - the 
signature or key record should _not_ be used by someone not aware of 
this tag. This seems quite possible. We could support this pretty easily 
by allowing marking of such critical tags, e.g. any tag beginning with 
c- is critical (and if not recognized, entire signature or key record 
should be ignored).

Nothing much.. Best, Amir Herzberg