{Found in an editor buffer - sorry if this is a bit late.}
> From: William Herrin <[email protected]>
> I'm suggesting that how we've been thinking about locators and
> identifiers is not quite right. The error was revealed by our attempt
> here to give them more precise definitions. They're eluding our
> attempts at rigor.
The problem is not that past thinking was lacking in rigor. (Please note that
the _original_ published definition of 'locator' was _extremely_ rigorous - it
meant a hierarchically structured name for a map node.)
The problem, _in part_, is that the terms, _over time_, have come to have
somewhat different meanings to different people. Hence my prior comment:
>> the best we can do for the Big Three (address, locator, identifier) is
>> to come up with definitions which reasonly match what they mean to
>> _most_ people now, and hope that the outliers eventually give up on
>> their idiosyncratic personal definitions ...
>> ... Note that the definitions for those three will have to be a little
>> 'loose', because if you make them _more_ specific, you will start to
>> loose people from the "_most_ people" pot above.
(I said 'in part' because _some_ people probably aren't thinking about this
rigorously.) But the point remains that much as you and I might want rigor,
that attempt will inevitably do a Titanic when faced with actual usage in a
wide community.
> Defining address in terms of where you are .. topologically [is not
> very plausible]
The term 'address', perhaps, for the reasons you give below (although I would
claim this doesn't apply to 'locator').
> ... offers a very slippery definition. It's hard to give it any kind of
> rigor in the face of anycast, multicast and load balancing.
Not to mention its use at the transport layer to name the entity at the other
end of a reliable end-end connection....
But you are absolutely right - 'address' is probably irretrievably broken.
The best we can do, probably, is say something akin to 'as used in deployed
IPv4', and then list its uses there.
Noel
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
