Re: [spfbis] Question about SPF checks based on RFC 7208

[Scott Kitterman <[e-mail]>]

On Sunday, May 01, 2016 09:24:24 PM Stuart Gathman wrote:
> On 04/30/2016 09:08 PM, Scott Kitterman wrote:
> > I think they are just wrong.  If either an IPv4 or IPv6 address is
> > returned it's not a void look up.  That's described at the end of 4.6.4
> > and is, I think clear.  I'd report it as a bug.
> If the connection is an IPv6 connection, then only AAAA RRs are queried
> (because an A RR could never match).  If the domain has only A RRs, then
> the AAAA queries will have empty results.   This seems to be
> indistinguishable from a non-existent domain.  This is the problem the
> OP is complaining about.
> 
> This means that pyspf, on detecting 2 "void" lookups for an AAAA, would
> then have to make at least 2 additional useless queries for the A record
> type to see if the lookups were actually "void" according to your
> interpretation.  Since I think your interpretation is somewhat
> reasonable, this *could* be an errata.
> 
> But I don't think so.  This is actually a case of "doctor, doctor, it
> hurts when I do this!"  Don't use an mx mechanism with A only names when
> valid connections can come in via IPv6.   Or at least, make it the
> *last* mechanism as recommended.  I can't think of any good reason to
> tweak or "clarify" the spec to enable this kind of SPF record.

I think it might make sense that to clarify that an mx mechanism is only void 
if neither A nor AAAA exist.  Implementations could either do the extra lookup 
or just not count it as void if only looking up one.  From an interoperability 
perspective, I think either would be fine.

Scott K