Hi Eric,

|>The strength of IP is that knowledge in the host is
|not _required_, 
|>but is allowed.  We're doing away with that capability.
|
|I wondered how you could agree with my point that hosts should not have
|visibility into the network by design and then state that capabilities
|to do so were removed from the host. It then hit me that my postings
|were about normal application data plane functions but you are
|apparently thinking about network management -- very different topics. 


Just to be very careful here: I agree that not requiring that the host know
about the network is a very strong idea.  However, I also feel pretty
strongly that the host should be able to know about the network.  Management
of the network (e.g., traceroute) is made possible only when the end-host
has visibility.


|I agree that managers cannot know anything about structures that occur
|at a different recursion layer than themselves. However, this is only
|one of a large repertoire of problems that make network management very
|challenging. The current state of network management for the Large End
|User is very troubling -- ditto for highly mobile networks. I've always
|viewed this as an inherent failing of network management itself (i.e.,
|the foundation is inadequate) but I concur that map and encaps makes it
|worse.


Is it really the failing of management?  Or is it concurrent with the
addition of a layer?  If you add a layer of abstraction, then how is
management supposed to work given that it is prohibited about knowing about
the lower layer?  In my book, this is exactly the architectural failing of
having an additional network layer.  Yes, we live with it today for the sake
of VPNs and the like, but is that really what we want for everyday life?

Regards,
Tony



_______________________________________________
rrg mailing list
[email protected]
https://www.irtf.org/mailman/listinfo/rrg

Reply via email to