43rd IETF in Orlando: Minutes of UDLR WG

[Emmanuel Duros <Emmanuel.Duros@sophia.inria.fr>]

I report here the minutes of UDLR working group held in Orlando (43rd
IETF meeting).

Emmanuel Duros
--------------

Presentation by Walid dabbous

I) Introduction

Presentation of a topology including an unidirectional link (udl). We
assume a small number of feeds compared to the number of receivers. As
an example a sat link with a 10th of feeds and a large set of receiver.


II) Summary of the document

The general idea is the emulation of a broadcast link over an
udl. Description of the possible scenarios which can occur on a
broadcast link -> the emulation must support all of them (See the draft
for the list of scenarios and the corresponding solutions). Once all
these scenarios are feasible on an udl we can then use routing
protocols.


III) Changes applied to the draft since version 01

1) In the early version: routing IP traffic from receivers to feeds via
the link layer tunnels was not allowed. After discussions with Aerospace
people and also with Harri Hakulinen (Nokia), we now allow routing for
flexibility reasons. As a result, encapsulating IP datagrams is no
longer performed according to the destination address of the IP datagram
but based on the next hop address (feed or receiver).

2) The paragraph concerning the hardware resolution protocol has been
modified. We do not address the hardware address resolution protocol on
long delay links, we consider it is out of scope in this working
group. However we propose to set the hardware address as a function of
the IP address of a receiver.


IV) Implementations

Single feed version of the link layer tunneling mechanism (LLTM)
implemented by WIDE:
 
  - succesful experiments performed with RIP, OSPF, DVMRP

Single/Multiple feed version implemented by INRIA:

  - Implemented as module independant of the hardware driver. 
  - Reliable Multicast experiment: W3C uses INRIA LLTM implementation
    to updates its mirror sites in Europe via satellite with MFTP.
    ETSI is also interested in this exp.
  - With Tunisian government: People from INRIA installed receivers in 
    Tunisia to bring fast access connectivity (Unicast/Multicast
    traffic). Note that no particular negociations with local ISP were
    required to route traffic (allowed by LLTM spec)
  - Testbed with Linux platform has been setup to simulate a multiple
    feed configuration -> LLTM implementation tested succesfuly
  - Code source available soon on the web page, announced on the ml.


V) Future status of the draft
(RC:Rob Coltun - WD: Walid Dabous)

Draft already available for more than 6 months on the ml. It is mature
enough to become an RFC: Informational or standard ?

RC: Is there anybody who has an objection to make it a standard ?
No objections.

Someone requests if there are some schedules about ongoing
standadisation/activity.

WD: Need interoperability tests with japan however satellite coverage
does not cover both regions. 

WD: In a multiple feed configuration, we will support several
implementations from various manufacturers. There are ongoing plans for
fast deployement in Europe and North Africa.

WD: Requested by RC, we will write a document which will describe
various experiments.

RC: This document should describe what has done in Multicast.

WD: Mrouted (DVMRP) works fine. We also have to see how PIM behaves over
the LLTM


VI) What next

RC: What is the future of the UDLR WG?

WD: We are going to focus on interoperability and standard track. Jon
Crowcroft sent a mail on a more generic infrastructure which has been
address yet.

Nghia Pham (Eutelsat): What about scalability ?

WD: We will see in the pilc BOF pb related to long delays which have
repercutions on protocol such as IGMP. We only deal here with routing
pb.


VII) Conclusion

Document will be sent for last call as a proposed standard RFC.