[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: updating GSE for the new millennium



On donderdag, mei 1, 2003, at 04:27 Europe/Amsterdam, Peter Tattam wrote:

I think only one relabelling should be allowed
Apart from the question if this is what we want, is this something we can enforce? Or should care to enforce? This would give ISPs the option of relabeling just so their customers can't do it themselves.

I would presume that a MH address be identifiable by just looking at the address (i.e. the top M (where M < N) bits represent a particular class of MH addresses).
Doing this has the advantage we can always easily recognize an unrewritten multihomed address, but is that a big enough advantage to require this? I can imagine a situation where hosts use a regular prefix as long as this prefix works and only rewrite with a secondary prefix when the first one fails.

The main issues I can see behind restoring the original packet would be TCP/UDP
checksums and IPSec. If the two end points are labelling aware, measures can
be taken to ignore the label from the checksums or replace it for IPsec,
otherwise it would be up to labelling boxen (border routers or other) to do
this on behalf of the hosts which are unaware of the labelling possibility.
Sure, if an end-host implements this itself it can optimize its internal processing to avoid doing unnecessary work. This is going to help even more with reachability detection.

One issue I just thought of is what happens if a physical site represents
multiple logical MH sites.
Then we're right back at the source address selection problem. :-(

However, in this case the source multihomed address would presumably always be chosen based on policy/application requirements rather than reachability status, so it shouldn't be as much of a problem.

For incoming traffic it's just a matter of keeping the mappings straight.

I wonder if these ideas might receive more support if the terminology of
labelling were used instead of GSE. We'd be more likely to draw in some
support from the MPLS mob.
It's probably too confusing. MPLS is all about very small labels that are attached and removed without touching the IP packet. What we're doing here involves globally unique values inside IP packets. I used the word "label" to avoid saying "address" or "name". We probably need something better for this.

Now using MPLS infrastructure to build a multihoming solution would be an interesting idea in it's own right.