RE: Some Comments on ID/Loc Separation Proposals

["Tony Li" <Tony.Li@procket.com>]

|  > From my perspective, we've been quite consistent with
|  > using locator for the topological naming for an interface, while
|  > an identifier is the name for a host.
|  
|  Ahhh...  This pretty much answers my question (I think :-)).
|  
|  In your taxonomy, the "ID" portion serves as a host/endpoint 
|  identifier and the "Locator" portion allows you to topologically 
|  locate a particular interface on the network, right?


Correct.  More specifically, we have not introduced any terminology
with respect to an 'endpoint'.

  
|  If I understand correctly,  the 8+8/GSE world uses a different 
|  breakdown between ID and Locator...  The upper 8 bytes
|  are only a topological locator for a particular link (which
|  may have many attached interfaces), and the lower order 8 bytes 
|  are used to identify a particular interface on that link.
|  
|  I consider this to be a fairly important architectural 
|  distinction, as the two models offer different levels of 
|  abstraction to ULPs.


Granted, however, from the macro point of view, this seems
somewhat less critical.  Our goal has been to "break as little
as possible" and towards that end, we have naturally gravitated
towards interface abstractions.  This is not a religious
requirement -- rather it's driven more by the pragmatics of
finding a solution that will gain widespread acceptance without
some very painful educational processes.

Tony