Minutes of
the Connectathon Mobile Ipv6
held
at Nancy
September 15-17 1999
Introduction
During a connectathon, some interoperability tests are achieved between different
implementations. In this context, a
connectathon related to the BGP4+ protocol took place in Grenoble, last year.
The purpose of the workshop so-called Connectathon
Mobile Ipv6 was focussed on another specific aspect of Ipv6: the mobility. The
Internet Mobility should allow all users to keep connected even when they move.
This includes the address getting and the movement detection in an automatic
way . The Mobile IP method supports transparency above the IP layer, that
means the
maintenance of active TCP connections and UDP port
bindings.
The Connectathon Mobile Ipv6 held at Nancy from the 15th
to the 17th of September 1999 in order to allow different developers
to test their mobility stack for the Ipv6 protocol and to validate the
interoperability of their software.
The tests event was organized by the G6[1]
group, which is the French group that supports the deployment of the Ipv6
protocol in France.
The detailed program is still available on the URL : http://www.loria.fr/~ichris/mobipv6.
The participants
The interoperability tests have been achieved between
four mobility stacks coming from different organizations: BULL (AIX) ,
ERICSSON-TELEBIT (FreeBSD) , NEC
(FreeBSD) and INRIA (FreeBSD).
List of participants :
|
NAME |
ORGANIZATION |
e-mail |
|
Ludovic BELLIER |
INRIA |
ludovic.bellier@inrialpes.fr |
|
Olivier
CHARLES |
FRANCE
TELECOM/CNET |
olivier.charles@cnet.francetelecom.fr |
|
Isabelle
CHRISMENT |
LORIA/Univ
Nancy2 |
isabelle.chrisment@loria.fr |
|
Laurent
CIARLETTA |
LORIA/NIST |
Laurent.ciarletta@loria.fr |
|
Jean-Michel
COMBES |
FRANCE
TELECOM/CNET |
jeanmichel.combes@cnet.francetelecom.fr |
|
Francis
DUPONT |
INRIA/GIE |
Francis.dupont@inria.fr |
|
Shuichi
KARINO |
NEC |
karino@ccs.mt.nec.co.jp |
|
Conny LARSSON |
ERICSSON |
Conny.Larsson@era-t.ericsson.se |
|
Aimé
LE ROUZIC |
BULL |
Aime.Le-Rouzic@bull.net |
|
Tsuyoshi
MOMOSE |
NEC |
t-momose@ccs.mt.nec.co.jp |
|
Simon NYBROE |
ERICSSON TELEBIT
|
sin@tbit.dk |
|
Mattias
PETTERSSON |
ERICSSON |
Mattias.Pettersson@era-t.ericsson.se |
|
Olivier ROBERT |
EUROCONTROL |
roberto@eurocontrol.fr |
|
Patrice ROMAND |
BULL |
Patrice.Romand@bull.net |
|
Hesham SOLIMAN |
ERICSSON |
Hesham.Soliman@era-t.ericsson.se |
|
Bertrand WALLRICH |
LORIA /INRIA |
bertrand.wallrich@loria.fr |
|
Sébastion
WOIRGARD |
Univ
. Strasbourg |
woirgard@dea-dess-info.u-strasbg.fr |
The tests event was very useful and meaningful for the
participants because they were confirmed that their implementation as
Correspondent, Home Agent or Mobile worked well with other implementations.
The test platform description
The test scenarios were done between a mobile node and a non-mobile correspondent
and also
between two mobiles.
Some tests have been run through the 6Bone, but the
major tests were achieved above Ethernet network over the local platform. A
more complete description of the test platform is given in the figure,
hereafter.
The movement was simulated by changing of IP subnets.
The Ethernet cables were unplugged and then plugged in another subnet .
The tests were considered as PASS when they worked
well with the following sequence:
- attach mobile node to home link.
2.
connect to correspondent via a telnet /ftp or a ping session.
- move to a foreign link
- check if the TCP connections are maintained or ping is always
running.
The checking was carried out by analysing the
exchanged Bindings during the ping or ftp/telnet session. Some protocol
analysers like tcpdump or nicedump[2] were used.
Some scenarios were also tested with mobiles supplied with WaveLAN cards. There
were two WaveLAN bases: one in the network recherche and another in the
network home (cf the platform figure).
home
|
home |
Suggestions for improvements in the draft
The different tests have shown that draft Mobility Support in IPv6 [3]should
be improved :
- When registering with the Home Agent, if no ACK was received and
the Binding Update had to be
retransmitted, a new sequence number should be sent (higher). It
was found that if the ACK was lost and the Mobile Node tries to resend the
Binding Update with the same sequence number then it will be rejected by the Home Agent (bad sequence
number).
- Although renumbering was not tested, the following case can be
foreseen: If the Mobile Node (being in the foreign network) receives a
Router advertisement with a new
prefix (renumbering), it may not necessarily choose the same
interface identifier. Therefore
when renumbering, the Home Agent should perform a DAD on the new address
before sending the Binding Ack. If that fails the Ack should be sent with
a rejection value in the status field. A new error code needs to be
created for this purpose (eg. "Invalid address"). This will remove the assumption that
the interface identifier is globally unique which may be a restriction.
- The draft has to be gone further into details about the Home Agent
Discovery.
Conclusion
The tests event was very successful thanks to the
participants. It has proved that now some Ipv6 mobility stacks exist and are
able to communicate together. Enough tests were done to show the basic
functions worked even some
implementations needs to be more stable.
But interoperability tests are not validity tests. We
tested only implementations
when communicating each other but not
the implementations behavior when illegal packets are received (except if a
implementation has some bugs and sends bad packets). Another style should be needed for improving a implementation.
At the end of the connectathon, A. Le Rouzic pointed
out that the movement detection has to
be clarified in terms of strategies between different links or movement, even
if today, Ipv6 provides enough mechanisms. The following questions were raised:
- How can we detect we have changes IP space ?
- Which level for the detection ?
- Level 3 (optimisation at level 2)
- How to do that ?
- Can NDP resolve the detection today ?
- Can we be sure the moving has been done ?
- Issues
- Default route changes
- Performances
- Border (How to choice the best router ?)
- Quality of Service
- Strategy to choice between several routers ?
During the connectathon, there was no test with IPSEC
and IKE. It seems the coexistence
between these protocols and Mobility IPv6 has to be
detailed within the draft. Friday morning , Francis Dupont did a presentation about Security and Mobility. He concluded that SA with mobile SHOULD use a wildcard source. Its slides are available on :
http://www.loria.fr/~ichris/mobipv6/connectathon99/fd_slides.htm