The Routing Over Large Clouds Mailing List Archive by date

[Curtis Villamizar <curtis@ans.net>]

In message <941203122001_4715026@aol.com>, Telford001@aol.com writes:
> 
> > Minutes of the Routing Over Large Clouds Working Group (ROLC)
> 
> 
> >Working Group Goals and Technical Problem Statement
> 
> >Analyze and propose enhancements to IP [and IPng] routing over large
> >Analyze and propose enhancements to IP [and IPng] routing over large
> >``shared media'' link layer (with respect to IP) networks (ATM, Frame
> >``shared media'' link layer (with respect to IP) networks (ATM, Frame
> >Relay, X.25, SMDS, etc.).
> 
> Wide-area links connected via wide-area MAC-bridges running a path
> Wide-area links connected via wide-area MAC-bridges running a path
> selection protocol like spanning tree or source routing can be used
> to build the switching fabric of a "Large Cloud" and should be added
> to the considerations of ROLC.

At prior meetings we agreed ti limit dicussions to scalable solutions
At prior meetings we agreed ti limit dicussions to scalable solutions
that could be made to work in an internetwork.  How do you propose
advertising the rest of the Internet to such a cloud?  The only
candidates are proxy ARP and running a routing protocol over the
candidates are proxy ARP and running a routing protocol over the
bridged cloud.  Neither scale at all.  Solutions that don't
internetwork at all are not up for consideration.

> >Avoid ``extra hops'' when routing IP. Extra hops are defined as points
> >Avoid ``extra hops'' when routing IP. Extra hops are defined as points
> >where an IP datagram leaves and re-enters the same link-layer-cloud as a
> >result of an IP routing decision, such as when routing between different
> >IP subnets overlaid on the same cloud.
> 
> Such a rule or recommendation is inappropriate.  This issue is a question
> of  internet design and not a question of protocol standardization.
> 
> It might make perfect sense to leave and reenter the "cloud" if selecting
> a path which remains within the cloud necessitates transversal of a link
> which has extremely low data rate or some other negative characteristic
> whereas the path which requires the external hop uses links of appropriate
> characterists.

First off, I think (hope) the intention is to allow but not require
you to avoid extra hops.  Second, this particular reference is to
receiving packets on one interface on a large cloud and sending it
back out the same interface.  This might only be appropriate in some
situations if there is buffering available and the cloud responds very
poorly to congestion.

> Joachim Carlo Santos Martillo Ajami
> Joachim Carlo Santos Martillo Ajami

Curtis