non-transitive cases (was: Purpose of terminology I-D)

[stepanek@electron.aero.org (James Stepanek)]

hi,

  IMHO, I find it both necessary and feasible to consider the
  non-transitive case.  In particular, handling this case without
  resorting to splitting up the satellite segment along
  unidirectional/ bidirectional lines implies avoiding an assumption
  of complete unidirectionality of all the links in the non-transitive
  segment.  I'm not convinced that any of the proposed solutions rely
  on such an assumption, but it is clearly worthy on analysis for the
  architecture and consideration in the terminology.

  Another groundstation-type which pops up at the DARPA BADD demos:
  sites which have no connectivity but for the unidirectional link.
  That is, that have no means of transmission back to the Loquacitor
  (my interim term for the feed/in-link).  Unfortunately, we can't do
  much for them.

have a nice day,

james


[On Fri, 25 Jul, Steve Deering writes:]
# At 3:22 AM -0700 7/25/97, Walid Dabbous wrote:
# > do you think that two way nodes attached to a non-trans
# > link require specific treatment? or is it just for
# > complete terminology purpose/
# 
# There exist satellite communication systems in which every groundstation
# has both send and receive capabilities, and which therefore act like a
# normal multi-access, bidirectional links.  There also exist satellite
# communication systems in which there are one or more send-only ground
# stations and many receive-only ground stations, e.g., DBS TV systems.
# I imagine that there are, or will be, hybrid systems, in which there
# are multiple groundstations capable of both send and receive, plus
# other groundstations that are receive-only (and perhaps also some send-
# only stations).  For communication between those nodes with send-and-rcv
# capability, the satellite channel is bidirectional, whereas for
# communication between other pairs of nodes, it is unidirectional.
# 
# For designing IP routing schemes for satellite networks, I thought it
# might be helpful to keep the more general non-transitive case in mind.
# In particular, a subset of nodes with both send-and-rcv capability can
# use the satellite channel to exchange routing information, rather than
# resorting to a terrestrial backchannel, and there might be good reasons
# to exploit that capability.
# 
# Alternatively, one could split such a hybrid network into two virtual
# networks, one unidirectional and one bidirectional, and ignore the more
# general non-transitive case.  But I suggest that it would be worth at
# least thinking about the pros and cons of those two different approaches.
# 
# Steve
# 
# 
#####



-- 
James Stepanek <stepanek@aero.org>   310/336-7911