wildcards and n-send NAPTRs

On 26 Jun 2008, at 10:05, Roy Arends wrote:

> As for prior art, there are many other examples where use of X and  
> wildcards are not recommended. DNAME and wildcard domain names. SRV  
> records and wildcard domain names.

Roy, just because a few specialised RRtypes can't be used with  
wildcards doesn't mean others should follow in that path. Just because  
others are doing something bad/stupid/dangerous can't be an excuse for  
the same behaviour. You wouldn't drink Coors Light (say) simply  
because lots of other people drink that awful liquid, would you? We  
both know why the semantics of DNAME and SRV (not forgetting NS and  
SOA) make wildcarding these RRtypes more than inappropriate. None of  
them have that property because of what's in their RDATA. I am *very*  
uncomfortable that n-send's impact on wildcarding is dependent on the  
record type's RATA and would have hoped you were too.

Even so, I can live with Ray's change of "MUST" to "SHOULD" here. It's  
still very ugly though.