Hi Noel,

>-----Original Message-----
>From: Noel Chiappa [mailto:[email protected]]
>Sent: Thursday, January 15, 2009 9:47 PM
>To: [email protected]
>Cc: [email protected]
>Subject: Re: [rrg] No liveness requirement in the ID/Loc Split concept
>
>
>    > From: Brian E Carpenter <[email protected]>
>
>    > I've said from day one that LISP's use of the term 'EID' is
incorrect,
>    > and it should be replaced by something like 'local locator' or
'level 0
>    > locator' or some such.
>
>Sorry, but it's _not_ a locator, and more an 'endpoint identifier' than
>anything else - at least, as far as the original definitions of those
terms
>go.
>
>(And might I remind people that the term "locator" came from Nimrod,
and "EID"
>from you-know-where. So excuse me if I get a little cranky with people
trying
>to tell me whether or not they are applicable to something, when I am
>perfectly darned happy with their use in that context.)
>
>The thing that people seem to be forgetting, in making this claim, is
that
>it's also important _what kind of thing_ is being named, not just _the
>attributes of the name itself_. Those are two orthagonal axes, and
people
>still don't seem to be clearly understanding/remembering that most
concept
>terms being thrown around here, such as locator, EID, etc, generally
involves
>a _set_ of choices, one on each axis.
>
>There seems to be a tendency to think 'locator' means 'name with
topological
>significance', and that's _not_ what it was defined as - it meant
'_interface
>name_ with topological significance'. (Actually, it was defined to be
>'interface name with topological significance which does not appear in
all
>packets', since people seemed terminally unable to wrap their minds
around the
>concept of an 'address' that didn't appear in all packets.) Similarly,
an
>'EID' is an '_endpoint_ identifier without topological significance',
not just
>'name with no topological significance'.
>
>We don't really have a generally-accepted term for 'endpoint name with
>topological significance'; 'address' has some of that flavour, but in
>e.g. IPv4 it also 'sort of' names an interface.
>
>
>With all this in hand, it's clear that a LISP EID does _not_ name an
>interface, but rather a 'stack' - i.e. an endpoint, with a collection
of TCP
>connections.

I think that depends on where you assign the EID (using "EID"
and "RLOC" in the LISP sense of the words for the purpose of
this message). If you assign the EID to an interface used
for forwarding packets to the outside world (e.g., an interface
attached to an ethernet link, a tunnel virtual interface, etc.)
then it names the interface in the traditional sense. If you
instead assign the EID to an internal virtual interface (e.g.,
a loopback interface), then in some sense it could be
considered as naming the stack. But...

> So to say that it's a 'locator' is wholly incorrect - it is
>_exactly_ an 'endpoint name', as that term was formally defined.

...it is still a routing locator at least within a limited scope.

LISP RLOCs are routing locators within the interdomain region,
and LISP EIDs appear in mapping tables within the interdomain
region. However, the EIDs are still routing locators within
edge networks even if the "edge network" consists of a
singleton end system.

IMHO, there is no escaping the fact that LISP EIDs are still
IP addresses that name interfaces, even though we have these
"special" interfaces like loopbacks that can in some sense
be considered as a "handle" for naming a stack.

HIP HITs on the other hand are true endpoint identifiers
that name end systems only, and cannot be used as routing
locators in any scope.

Fred
[email protected]
 
>There is no one-word description/term which _exactly_ describes a LISP
EID;
>it's a bit of a kludge (precisely because of installed base issues). To
be
>maximally precise, a LISP 'EID' is 'a globally-unique endpoint name
with
>topological significance within a local scope only'.
>
>       Noel
>_______________________________________________
>rrg mailing list
>[email protected]
>http://www.irtf.org/mailman/listinfo/rrg
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to