On 06/03/2016 06:55 PM, Tom Lane wrote:
> More importantly,
> on what basis do you conclude that the inet type will only be asked to
> store link-local addresses that are currently valid on the local machine?
> It is not very hard to think of applications where that wouldn't be the
> case.
That's a good point.
> I think a better plan is just to store the zone string verbatim. It is
> not our job to check its validity or try to determine which spellings
> are equivalent.
Agreed.
That leaves me wondering if it's really worth extending INET, though.
TEXT would be just fine to store a textual scope id. And it makes it
utterly clear that there's no magic involved.
Kind Regards
Markus Wanner