>>>>> "Vishwas" == Vishwas Manral <[email protected]> writes:
    Vishwas> Hi Michael,
    Vishwas> I am working on updating the text of the draft. I had a few
    Vishwas> doubts.

(fixing the quoting to clarify who said what)

    mcr> I think the focus on "movement" and "mobility" is wrong.
    mcr> Instead, focus on connectivity.
    mcr> 
    mcr> Tero (and I) asked the question about a device which was, from a
    mcr> topological point of view, originally connected to "outside" (on the
    mcr> public Internet, effectively), and had built up a mesh.
 
    Vishwas> Are you talking of a case where the device had Public internet
    Vishwas> connectivity and has now roamed to a place where it is inside the 
network,
    Vishwas> so the IPsec tunnel is not required. You are probably looking at a 
way to
    Vishwas> gracefully tear down the IPsec tunnel and use the native 
connectivity. Is
    Vishwas> the understanding of the usecase you had in mind correct?

Yes, this is an example, but not the only situation.

"Public internet connectivity" is not exactly the right term, because
that almost implies that the device was never NAT'ed, but in fact it
might have been.  What matters is that it wasn't behind the home
enterprises' security gateway(s).

    mcr> There is a connectivity change (let's not worry about how, why, but I
    mcr> can describe a dozen ways if you like, and not all of them involve 3G 
or
    mcr> 802.11), and the device is now connected behind a gateway which is also
    mcr> part of the mesh.
    mcr> 
    mcr> Let's assume a multiple star topology situation. (We draw these on
    mcr> napkins in Paris. I promise to re-read the problem statment to learn if
    mcr> this scenario has been named)
    mcr> 
    mcr> Let's say that said device ("A") previously was part of 
VPN-Concentrator X.
    mcr> (It was perhaps part of that star via NAT-T transport).  Packets flowed
    mcr> from A to device B.  Device B is located on network B-location, and is
    mcr> served by B-gateway, and these packets flowed via VPN-Concentrator X.
    mcr> 
    mcr> X attempts to ask A to connect directly to B-gateway so that packets
    mcr> from A<->B take the shortcut.  A presently has connectivity *on*
    mcr> "network B-location".  That is, A and B are in fact, co-located (but
    mcr> might not be on the same link).
    mcr> 
    mcr> 1) Is there a requirement that B-gateway must be able to make a bypass
    mcr> connection to another spoke, from *behind* B?

    Vishwas> Let me see if I understood the question. We are talking of a 
gateway-B
    Vishwas> which is on the path between 2 spokes (A and B) which are talking 
via a VPN
    Vishwas> concentrator (X). You are asking if we allow the Gateway-B to 
allow for a
    Vishwas> bypass and if so can we allow for a "clear" bypass connectivity. 
Is my
    Vishwas> understanding correct?

Yes.
Just to be clear, A and B are *hosts*
A was a "road warrior".  Whether or not B is also a road warrior is not
relevant, what's important is that B is behind B-gateway, and gets VPN
service from B-gateway (and concentrator-X).

    Vishwas> I agree there is a use case for this optimization, but do we want 
to tackle
    Vishwas> the use case as part of the base solution or as an extension once 
the base
    Vishwas> solution is created (so a future use case).

good enough.

-- 
Michael Richardson <[email protected]>, Sandelman Software Works 
IETF ROLL WG co-chair.    http://datatracker.ietf.org/wg/roll/charter/

Attachment: pgpAXAtVzjpcL.pgp
Description: PGP signature

_______________________________________________
IPsec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ipsec

Reply via email to