"Meng Weng Wong" asked:
Should we change the spec to say that if the argument to an
"a" mechanism is obviously an IP address, we interpret it as
ip4 / ip6?
I vote 'No' to all undocumented tolerations.
The danger is that someone tests against an MTA implementation which accepts
more than the spec allows, then wonders why it does not work with other
implementations.
Either put things like ipv4 == ip4 == a in the spec, or don't permit them at
all.
Putting alternatives in the spec is pretty stupid (unless for an engineering
purpose, such as reducing the total character count). Why, if someone is
reading the spec to write the record, should they need a choice? And if they are
not reading the spec. that's their own look-out.
My recommendation: implement only what is in the spec, and don't put unnecessary
alternatives in the spec.
Chris Haynes