Excerpts from Noel Chiappa on Mon, Mar 30, 2009 05:43:02PM -0400:
> > From: Scott Brim <[email protected]>
>
> > Is "the place you're trying to end up at" a point in the
> > topology where something is attached, or is it the something
> > itself?
>
> The former, although presumably one only wants to get to the former
> in order to get to the latter.
OK. It was a layering question. "you" are apparently the network
trying to deliver a packet to an attachmen point, as opposed to an
endpoint trying to deliver a packet to a peer endpoint. I would have
gone the other way but now I know what you mean.
> > If it's the second, you might call the name of it an
> > identifier.
>
> Normally, yes.
>
> However, if the name of the latter (for whatever reason) had
> information about the location built into it (think of them as
> 'locators for endpoints'), then it would not be an identifier. Now,
> I happen to think that's a bad approach, in engineering terms (q.v.
> IPvN address), but it is logically consistent (in terms of the
> terminology).
OK
> > a name used as a locator can have any amount of structure in it, more
> > or less, and that doesn't change whether it's a locator or not.
>
> That's kind of like being 'a little bit pregnant'. A name either has _no_
> location information _built into it_, or is has some - and all the latter
> such names are 'locators'. (And notice that IEEE numbers have _no_ location
> information built into them.)
OK
> > I'm saying that the degree of structure in the name does not decide
> > whether it can be called a locator or not.
>
> Under the above definition, I concur.
>
> >> To most of us, "locator" means 'a _structured_ name for a place in the
> >> topology'.
>
> > Does it matter to you how much structure is in the name?
>
> Not as long as the structure is related to the location. For a
> counter-example, even though an IEEE number does have stucture, that
> structure is not about the location.
OK. Those two almost go hand-in-hand, but not quite. I want to
assert that that attribute:
- A locator has structure in it related to topological position
is a subset of the current working definition:
- A locator is associated with topological position such that if a
node moves its locator may have to change.
The reason is that the second includes the first but can be true even
if a name does not have any structure in it. Test case: suppose I am
an evil network administrator and I decide that all odd-numbered IP
addresses need to be on that Ethernet over there, and all
even-numbered IP addresses on this Ethernet over here. At least for
the final steps, there is no structure related to location, and yet it
may have to change if a node moves between those two Ethernets.
Thoughts?
> >> the _original_ definition of "locator" did _not_ include use
> >> "by [the] forwarding": the term 'locator' was defined
> >> _precisely_ to have a term that meant 'structured name for a
> >> place in the topology that is _not_ used by the forwarding'.
>
> > How was it used then?
>
> To name map nodes.
>
> > What did forwarding use?
>
> Flow identifiers (think MPLS tags).
OK
Scott
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
