more tunnels and what to do next

[Pedro Roque Marques]

>>>>> "Quaizar" == Quaizar Vohra <qv@cs.unh.edu> writes:

    Quaizar> Hi Jim,
    >>  What I think we need to do after configuring this with UNH on
    >> the east coast.  Is determine a way of automating prefix
    >> distribution on the 6bone with the tunnel end points.  You
    >> should be able to directly send to UNH which is our leg of the
    >> 6bone on the U.S. East coast. And not have to go through West
    >> Coast given the prefix of the node based on RFC 1897.  We
    >> should have the East Coast end point up soon.
    >> 
    >> If we could develop an algorithm to generate the IPv4-Tunnel
    >> end point from the prefix which may be possible using RFC 1897
    >> that would help a lot.
    >> 
    >> /jim
    >> 

    Quaizar> Looks like a good idea. How about everyone having 24 bit
    Quaizar> IPv4 subnets. so that we have 64 bit prefixes formed from
    Quaizar> there. Ex 132.177.118.0 is 84b1:7600 when I form a prefix
    Quaizar> I get 5f02:3000:7600:84b1::/64. Then all can agree that
    Quaizar> some standard last octet can be used to find the v4
    Quaizar> address for the tunnel endpoint. How abot using the last
    Quaizar> octet of your autonomous system # e.g. for us ASN is
    Quaizar> 0x0230 so the last octet is 0x30 so the the tunnel
    Quaizar> endpoint is 132.177.118.48.

    Quaizar> This is pretty restrictive though, may be something on
    Quaizar> similar lines.

Sorry, but the idea seams terrible.

This way to configure a tunnel you are requiring that end points have
a particular IPv4 address which might be already in use by another
system on your network, no matter what the scheme is. Also you impose
that the is only one tunnel end-point per prefix. Some people already
have more than one (for different tunnels of course), if i'm not
mistaken, and i was planning on doing  the same.

The second point is that i really don't understand what you're trying
to achieve. If we're talking about static configuration here, then all
you need to know is the other end prefix and v4 end-point. As you've
seen the end-point info doesn't add too much space to the prefix list
that you need anyway.

And, as things are today, you can configure unidirectional
links, i.e. configure a tunnel for which the other end point knows
nothing, if you have the prefix and v4 address. Your proposal makes
this even easier to happen, which i don't think is a good idea.

If you ask the DNS, it will happily tell you quad-A records for my
network but i would apreciate that people wouldn't start dumping
packets to my tunnel end-point without prior agreement.

The original question is if packets from A should go to B before
reaching C, C being closer to A. I believe the answer to be: set a
tunnel between A and C and tell B about it.

regards,
  Pedro.