In einer eMail vom 04.06.2010 19:31:32 Westeuropäische Sommerzeit schreibt  
[email protected]:

On 4  June 2010 at 16:41:15 Tony Li sent:
> 
> > But RRG's charter  includes exploration of "addressing" problems.
> > An "address" is  used to determine forwarding operation.
> > So it's natural to  discuss forwarding issues here. Right?
> 
> 
> Ummm...   No, not in the full generality of how some people use that term.
>  
> Again, we were chartered to discuss routing and addressing  architecture.
> Forwarding issues typically involve interesting  questions about how you 
can
> structure your FIB to minimize painful  memory accesses, what type of 
memory
> to use, how it should be  interconnected to your processing element, how 
to
> deal with ECMP,  dealing with indirect next hops and the like.  All of 
those
> are  out of scope.
> 
> Regards,
> Tony
> 
> 
>  

OK, I agree.
Tony, 
however I do disagree. You just  looked at what rtgwg is doing  (ECMP, 
"via"-routes ) and concluded such topics belong to rtgwg, hence not to  RRG, 
hence is out of scope.
 
First:the IAB-report talks a lot about Moore's law.
Second: A routing architecture is more than just solving the scalability  
problem.
Obviously you only look backwards: LISP is (as said by their authors  
themselves) a jackup solution. ILNP revitalizes ideas which exist at least  for 
one decade (right? btw this is not a  disadvantage per se. But it  is  
nothing new under the sun; it is still the wrong way of understanding  
hierarchy )
 
A routing architecture is more. It should concurrently serve all the  
important issues:
-scalability (of course)
-mobility, multihoming,TE forward (knowing multipath alternatives) and  
backward (figuring out to choose the right multipath alternative) e.g. for  the 
sake of traffic balancing or the sake of congestion handling.
-multicast ( I don't know what is more disappointing: the IETF multicast  
technology or the fact that multicast is a non-issue in the RRG).
 
By addressing one issue after the other you will get any arbitrary  
solution in the end which is cast by the sequence these issues have been  
addressed. (btw, this is my general observation wrt to all IETF progresses,  
too)
 
Heiner
 
Heiner
  


But  there is one certain point of intersection of addressing, forwarding 
and path  selection:
As clear an address shows a node's topological location, that  much it 
facilitates hop by hop path selection to that node, says: decision  making for 
forwarding, sequentially.

So why not just embed next hops in  an address: locator?
_______________________________________________
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