Re: Questions about tunneling proposals

[Stephen Schwab <schwab@twinsun.com>]

I'll pipe up...

ygz> Yes they are close, and so is our virtual symmetric network approach
ygz> in HRL.

Everyone considers the virtual network an elegant solution.  The devil
is in the details...

te> In both proposals there is a virtual reverse link to elude routing
te> protocols as if there were a bidirectional satellite link.  The real
te> difference I can see is that Aerospace is going deeper in the level of
te> abstraction.  Only routing traffic is using this virtual reverse satellite
te> link in Wide's proposal.  What about Aerospace?  I guess all traffic from 
te> receivers to feeds isn't encapsultad, but I don't see this point anywhere
te> clearly addressed in the draft.  Could you have a word about that?

ygz> Agree.  It should be explicitly mentioned in the draft that the tunnel
ygz> is used for routing protocol messages only.

This is not the case for the VIPRe implementation.  VIPRe provides a tunnel
to carry encapsulated IP packets.  We don't restrict to RIP or OSPF packets.
(James, please correct me if the implementation has changed recently to
restrict which packets traverse the reach-back network as encapsulated vs.
non-encapsulated packets.)

Once the existing routing implementations see a bi-directional network,
albeit a virtual one, they work, in that they provide connectivity.
In theory only the routing protocols should need to use the virtual network.

For the environment in which VIPRe was used, we wanted to maximize the 
likelihood that all the existing software would work, even if things were
configured strangely.  And we wanted to minimize the changes needed on the
hosts connected to satellite receivers.  Using the virtual network for
all traffic is probably overkill, but it seemed the best way to make things
work initially.  

How many of those reading this list have actually used their implementations
over a satellite link?  (Or a one-way cable distribution tree?)  

Do things work fine once routing works, or are there other
configuration gotchas on the hosts?  How many applications and/or services
have actually run over these networks?

--Steve

p.s. My return address is schwab@twinsun.com.  Mail is forwarded from
schwab@aero.org.