INTERNET DRAFT R. Boivie, N. Feldman
<draft-ooms-xcast-basic-spec-02.txt> IBM
Y. Imai
Fujitsu
W. Livens
Colt Telecom
D. Ooms, O. Paridaens
Alcatel
October, 2001
Expires April, 2002
Explicit Multicast (Xcast) Basic Specification
Status of this Memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as Internet-
Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
Abstract
Multicast has become increasingly important with the emergence of
network-based applications such as IP telephony and video
conferencing. The Internet community has done a significant amount of
work on IP multicast over the last decade [1075, 2201, 2236, DEER,
DEE2, FARI, HOLB, MBONE, PERL].
However, while traditional multicast schemes are scalable in the
sense that they can support very large multicast groups, there are
scalability issues when a network needs to support a very large
number of distinct multicast groups.
Ooms, et al. Expires April 2002 [Page 1]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
This document describes a new scheme for multicast - named Explicit
Multicast (Xcast) - that complements the existing schemes. Whereas
the existing schemes can support a limited number of very large
multicast sessions, Xcast can support a very large number of small
multicast sessions. This is achieved by explicitly encoding the list
of destinations in the data packets, instead of using a multicast
address.
The co-operation of Xcast - a data plane mechanism - with existing
control planes is described and scenarios for gradual deployment are
investigated. Encodings for both IPv4 and IPv6 are proposed.
This draft merges three earlier drafts ([BOIV], [IMAI], [OOMS]).
Table of Contents
1. Introduction
2. Xcast Overview
3. The cost of the traditional multicast schemes
4. Motivation
5. Application
6. Xcast Flexibility
7. Control plane
7.1. SIP
7.2. Receiver Initiated Join model
8. Optional information
8.1. List of ports
8.2. List of DSCPs
8.3. Channel Identifier
9. Encoding
9.1. General
9.2. IPv4
9.2.1. IPv4 header
9.2.2. Xcast4 header
9.3. IPv6
9.3.1. IPv6 header
9.3.2. Xcast6 header
9.3.2.1. Routing Extension header
9.3.2.2. Destination Extension header
10. Impact on Upper Layer Protocols
10.1. Checksum calculation in UDP and ICMP
10.2. IPsec Authentication header
11. Gradual Deployment
11.1. Tunneling
11.2. Premature X2U
11.3. Semi-permeable tunneling (IPv6 only)
11.4. Special case: deployment without network support
Ooms, et al. Expires April 2002 [Page 2]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
12. (Socket) API
13. Security Considerations
References
Changes:
00->01: - added Channel Identifier in Xcast header
- cleaned up wording (traditional, current, today's mc)
- added definition of Session and Channel
01->02: - Channel Identifier in the Xcast4 header was moved higher
- editorial changes
1. Introduction
Multicast, the ability to efficiently send data to a group of
destinations, is becoming increasingly important for applications
such as IP telephony and video-conferencing.
There seem to be two kinds of multicast that are important: a
broadcast-like multicast that sends data to a very large number of
destinations and a "narrowcast" multicast that sends data to a fairly
small group. An example of the first is the audio and video
multicasting of a presentation to all employees in a corporate
intranet. An example of the second is a videoconference involving 3
or 4 parties. For reasons described below, it seems prudent to use
different mechanisms for these two cases. As the reliable multicast
transport group has stated: "it is believed that a 'one size fits
all' protocol will be unable to meet the requirements of all
applications" [RMT]. Note too, that [2902], "Overview of the 1998 IAB
Routing Workshop", also came to the same conclusion: "For example,
providing for many groups of small conferences (a small number of
widely dispersed people) with global topological scope scales badly
given the current multicast model".
Today's multicast schemes can be used to minimize bandwidth
consumption. Explicit Multicast (Xcast) also can be used to minimize
bandwidth consumption for "small groups". But it has an additional
advantage as well. Xcast eliminates the per session signaling and
per session state information of traditional multicast schemes and
this allows Xcast to support very large numbers of multicast
sessions. And this scalability is important since it enables
important classes of applications such as IP telephony,
videoconferencing, collaborative applications, networked games etc.
where there are typically very large numbers of small multicast
groups.
2. Xcast Overview
Ooms, et al. Expires April 2002 [Page 3]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
In this document the following terminology will be used:
- Session: in Xcast the term 'multicast session' will be used instead
of 'multicast group' to avoid the strong association of multicast
groups with multicast group addresses in traditional IP multicast.
- Channel: in a session with multiple senders (e.g. a video
conference), the flow sourced by one sender will be called a channel.
So a session can contain one or more channels.
In the Host Group Model the packet carries a multicast address as a
logical identifier of all group members. In Xcast, the source node
keeps track of the destinations in the multicast channel that it
wants to send packets to.
The source encodes the list of destinations in the Xcast header, and
then sends the packet to a router. Each router along the way parses
the header, partitions the destinations based on each destination's
next hop, and forwards a packet with an appropriate Xcast header to
each of the next hops.
When there is only one destination left, the Xcast packet can be
converted into a normal unicast packet, which can be unicasted along
the remainder of the route. This is called X2U (Xcast to Unicast).
For example, suppose that A is trying to get packets distributed to
B, C & D in Figure 1 below:
R4 ---- B
/
/
A----- R1 ---- R2 ---- R3 R8 ---- C
\ /
\ /
R5 ---- R6 ---- R7
\
\
R9 ---- D
Figure 1
This is accomplished as follows: A sends an Xcast packet with the
list of destinations in its Xcast header to the first router, R1.
Since the Xcast header will be slightly different for IPv4 and IPv6
we won't reveal any details on the encoding of the Xcast header in
this section (see section 9). So, ignoring the details, the packet
that A sends to R1 looks like this:
Ooms, et al. Expires April 2002 [Page 4]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
[ src = A | dest = B C D | payload ]
When R1 receives this packet, it needs to properly process the Xcast
header. The processing that a router does on receiving one of these
Xcast packets is as follows:
- Perform a route table lookup to determine the next hop for each of
the destinations listed in the packet.
- Partition the set of destinations based on their next hops.
- Replicate the packet so that there's one copy of the packet for
each of the next hops found in the previous steps.
- Modify the list of destinations in each of the copies so that the
list in the copy for a given next hop includes just the destinations
that ought to be routed through that next hop.
- Send the modified copies of the packet on to the next hops.
- Optimization: If there is only one destination for a particular
next hop, send the packet as a standard unicast packet to the
destination (X2U), since there is no advantage to forwarding it as an
Xcast packet.
So, in the example above, R1 will send a single packet on to R2 with
a destination list of < B C D > and R2 will send a single packet to
R3 with the same destination list.
When R3 receives the packet, it will, by the algorithm above, send
one copy of the packet to next hop R5 with an Xcast list of < C D >,
and one ordinary unicast packet addressed to < B > to R4. R4 will
receive a standard unicast packet and forward it on to < B >. R5 will
forward the Xcast packet that it receives on to R6 which will pass it
on to R7. When the packet reaches R7, R7 will transmit ordinary
unicast packets addressed to < C > and < D > respectively. R8 and R9
will receive standard unicast packets, and forward the packets on to
< C > and < D > respectively.
It's important that the Xcast packet that is sent to a given next hop
only includes destinations for which that next hop is the next hop
listed in the route table. If the list of destinations in the packet
sent to R4, for example, had also included C and D, R4 would send
duplicate packets.
Note that when routing topology changes, the routing for an Xcast
channel will automatically adapt to the new topology since the path
an Xcast packet takes to a given destination always follows the
Ooms, et al. Expires April 2002 [Page 5]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
ordinary, unicast routing for that destination.
3. The cost of the traditional multicast schemes
Traditional multicast schemes [DEER, DEE2, FARI] were designed to
handle very large multicast groups. These work well if one is trying
to distribute broadcast-like channels all around the world but they
have scalability problems when there is a very large number of
groups.
The characteristics of the traditional IP multicast model are
determined by its two components: the Host Group model [DEER] and a
Multicast Routing Protocol. Both components make multicast very
different from unicast.
In the Host Group model, a group of hosts is identified by a
multicast group address, which is used both for subscriptions and
forwarding. This model has two main costs:
- Multicast address allocation: The creator of a multicast group
must allocate a multicast address which is unique in its scope
(scope will often be global). This issue is being addressed by
the Malloc working group, which is proposing a set of Multicast
Address Allocation Servers (MAAS) and three protocols (MASC, AAP,
MADCAP).
- Destination unawareness: When a multicast packet arrives in a
router, the router can determine the next hops for the packet, but
knows nothing about the ultimate destinations of the packet, nor
about how many times the packet will be duplicated later on in the
network. This complicates the security, accounting and policy
functions.
In addition to the Host Group model, a routing algorithm is required
to maintain the member state and the delivery tree. This can be done
using a (truncated) broadcast algorithm or a multicast algorithm
[DEER]. Since the former consumes too much bandwidth by
unnecessarily forwarding packets to some routers, only the multicast
algorithms are considered. These multicast routing protocols have
the following costs:
- Connection state: The multicast routing protocols exchange
messages that create state for each (source, multicast group) in
all the routers that are part of the point-to-multipoint tree.
This can be viewed as "per flow" signaling that creates multicast
connection state, possibly yielding huge multicast forwarding
tables. Some of these schemes even disseminate this multicast
Ooms, et al. Expires April 2002 [Page 6]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
routing information to places where it isn't necessarily needed
[1075]. Other schemes try to limit the amount of multicast
routing information that needs to be disseminated, processed and
stored throughout the network. These schemes (e.g. [2201]) use a
"shared distribution tree" that is shared by all the members of a
multicast group and they try to limit the distribution of
multicast routing information to just those nodes that "really
need it". But these schemes also have problems. Because of the
shared tree, they use less than optimal paths in routing packets
to their destinations and they tend to concentrate traffic in
small portions of a network. And these schemes still involve lots
of "per flow" signaling and "per flow" state.
- Source advertisement mechanism: Multicast routing protocols
provide a mechanism by which members get 'connected' to the
sources for a certain group without knowing the sources
themselves. In sparse-mode protocols [2201, DEE2], this is
achieved by having a core node, which needs to be advertised in
the complete domain. On the other hand, in dense-mode protocols
[1075] this is achieved by a "flood and prune" mechanism. Both
approaches raise additional scalability issues.
- Interdomain routing: Multicast routing protocols that rely on a
core node [2201, DEE2] additionally need an interdomain multicast
routing protocol (e.g. [FARI]).
The cost of multicast address allocation, destination unawareness and
the above scalability issues lead to a search for other multicast
schemes. Source-Specific Multicast (SSM) [HOLB] addresses some of
the above drawbacks: in SSM a host joins a specific source, thus the
channel is identified by the couple (source address, multicast
address). This approach avoids multicast address allocation as well
as the need for an interdomain routing protocol. The source
advertisement is taken out of the multicast routing protocol and is
moved to an out-of-band mechanism (e.g. web page).
Note that SSM still creates state and signaling per multicast channel
in each on-tree node. Figure 2 depicts the above costs as a function
of the number of members in the session or channel. All the costs
have a hyperbolic behavior.
Ooms, et al. Expires April 2002 [Page 7]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
cost of the traditional
multicast model
per member
^
| costly| OK
| <-----|----->
| . |
| .. |
| ..|..
| | .........
| | ........
+--------------------------->
| number of members
v
alternative=Xcast
Figure 2
The traditional multicast model becomes expensive for its members if
the groups are small. Small groups are typical for conferencing,
gaming and collaborative applications. These applications are well-
served by Xcast.
In practice, traditional multicast routing protocols impose
limitations on the number of groups and the size of the network in
which they are deployed. For Xcast these limitations do not exist.
4. Motivation
Xcast takes advantage of one of the fundamental tenets of the
Internet "philosophy", namely that one should move complexity to the
edges of the network and keep the middle of the network simple. This
is the principle that guided the design of IP and TCP and it's the
principle that has made the incredible growth of the Internet
possible. For example, one reason that the Internet has been able to
scale so well is that the routers in the core of the network deal
with large CIDR blocks as opposed to individual hosts or individual
"connections". The routers in the core don't need to keep track of
the individual TCP connections that are passing through them.
Similarly, the IETF's diffserv effort is based on the idea that the
routers shouldn't have to keep track of a large number of individual
RSVP flows that might be passing through them. It's the authors'
belief that the routers in the core shouldn't have to keep track of a
large number of individual multicast flows either.
Compared to traditional multicast, Xcast has the following
advantages:
Ooms, et al. Expires April 2002 [Page 8]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
1) Routers do not have to maintain state per session (or per channel)
[SOLA]. This makes Xcast very scalable in terms of the number of
sessions that can be supported since the nodes in the network do not
need to disseminate or store any multicast routing information for
these sessions.
2) No multicast address allocation required.
3) No need for multicast routing protocols (neither intra- nor
interdomain). Xcast packets always take the "right" path as
determined by the ordinary unicast routing protocols.
4) No core node, so no single point of failure. Unlike the shared
tree schemes, Xcast minimizes network latency and maximizes network
"efficiency".
5) Symmetric paths are not required. Traditional multicast routing
protocols create non-shortest-path trees if paths are not symmetric.
(A path between two nodes A and B is symmetric if the path is both
the shortest path from A to B as well as the shortest path from B to
A.) It is expected that an increasing number of paths in the
Internet will be asymmetric in the future as a result of traffic
engineering and policy routing, and thus the traditional multicast
schemes will result in an increasing amount of suboptimal routing.
6) Automatic reaction to unicast reroutes. Xcast will react
immediately to unicast route changes. In traditional multicast
routing protocols a communication between the unicast and the
multicast routing protocol needs to be established. In many
implementations this is on a polling basis, yielding a slower
reaction to e.g. link failures. It may also take some time for
traditional multicast routing protocols to fix things up if there is
a large number of groups that need to be fixed.
7) Easy security and accounting. In contrast with the Host Group
Model, in Xcast all the sources know the members of the multicast
channel, which gives the sources the means to e.g. reject certain
members or count the traffic going to certain members quite easily.
Not only a source, but also a border router is able to determine how
many times a packet will be duplicated in its domain. It also
becomes easier to restrict the number of senders or the bandwidth per
sender.
8) Heterogeneous receivers. Besides the list of destinations, the
packet could (optionally) also contain a list of DiffServ CodePoints
(DSCPs). While traditional multicast protocols have to create
separate groups for each service class, Xcast incorporates the
possibility of having receivers with different service requirements
Ooms, et al. Expires April 2002 [Page 9]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
within one multicast channel.
9) Xcast packets can make use of traffic engineered unicast paths.
10) Simple implementation of reliable protocols on top of Xcast,
because Xcast can easily address a subset of the original list of
destinations to do a retransmission.
11) Flexibility (see section 6).
12) Easy transition mechanisms (see section 11).
It should be noted that Xcast has a number of disadvantages as well:
1) Overhead. Each packet contains all remaining destinations. But,
the total amount of data is still much less than for unicast (payload
is only sent once). A method to compress the list of destination
addresses might be useful.
2) More complex header processing. Each destination in the packet
needs a routing table lookup. So an Xcast packet with n destinations
requires the same number of routing table lookups as n unicast
headers. Additionally, a different header has to be constructed per
next hop. Note however that:
a) Since Xcast will typically be used for super-sparse sessions,
there will be a limited number of branching points, compared to
non-branching points. Only in a branching point do new headers
need to be constructed.
b) The header construction can be reduced to a very simple
operation: overwriting a bitmap.
c) Among the non-branching points, a lot of them will contain only
one destination. In these cases normal unicast forwarding can be
applied.
d) By using a hierarchical encoding of the list of destinations in
combination with the aggregation in the forwarding tables the
forwarding can be accelerated ([OOMS]).
e) When the packet enters a region of the network where link
bandwidth is not an issue anymore, the packet can be transformed
by a Premature X2U. Premature X2U (see section 11.2) occurs when
a router decides to transform the Xcast packet for one or more
destinations into unicast packets. This avoids more complex
processing downstream.
Ooms, et al. Expires April 2002 [Page 10]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
f) Other mechanisms to reduce the processing have been described
in [IMAI] (tractable list) and [OOMS] (caching), but are not (yet)
part of this basic Xcast specification.
3) Xcast only works with a limited number of receivers.
5. Application
While Xcast is not suitable for multicast sessions with a large
number of members, such as the broadcast of an IETF meeting, it does
provide an important complement to existing multicast schemes in that
it can support very large numbers of small sessions. Thus Xcast
enables important applications such as IP telephony,
videoconferencing, multi-player games, collaborative e-meetings etc.
The number of these sessions will become huge.
Some may argue that it is not worthwhile to use multicast for
sessions with a limited number of members, and use unicast instead.
But in some cases limited bandwidth in the "last mile" makes it
important to have some form of multicast as the following example
illustrates. Assume n residential users that set up a video
conference. Typically access technologies are asymmetric (e.g. xDSL,
GPRS or cable modem). So, a host with xDSL has no problem receiving
n-1 basic 100kb/s video channels, but the host is not able to send
its own video data n-1 times at this rate. Because of the limited
and often asymmetric access capacity, some type of multicast is
mandatory.
A simple but important application of Xcast lies in bridging the
access link. The host sends the Xcast packet with the list of
unicast addresses and the first router performs a Premature X2U.
Since Xcast is not suitable for large groups, Xcast will not replace
the traditional multicast model, but it does offer an alternative for
multipoint-to-multipoint communications when there can be very large
numbers of small sessions.
6. Xcast Flexibility
The main goal of multicast is to avoid duplicate information flowing
over the same link. By using traditional multicast instead of
unicast, bandwidth consumption decreases while the state and
signaling per session increases. Xcast has a cost of 0 in these 2
dimensions, but it does introduce a third dimension corresponding to
the header processing per packet. This three dimensional space is
depicted in Figure 3.
Ooms, et al. Expires April 2002 [Page 11]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
state&signaling
per session
in router
^
|
|
....
B | ....
. | ....
. | ....
. | ....
. +------------------..---> processing
. / .... C per packet
. / ..... in router
. / .....
. / .....
./ .....
/A....
/
/
link bandwidth
Figure 3
One method of delivering identical information from a source to n
destinations is to unicast the information n times (A in Figure 3).
A second method, the traditional multicast model (B in Figure 3)
sends the information only once to a multicast address. In Xcast the
information is sent only once, but the packet contains a list of
destinations (point C).
The three points A, B and C define a plane (indicated with dots in
Figure 3): a plane of conservation of misery. All three approaches
have disadvantages. The link bandwidth is a scarce resource,
especially in access networks. State&signaling/session encounters
limitations when the number of sessions becomes large and an
increased processing/packet is cumbersome for high link speeds.
One advantage of Xcast is that it allows a router to move within this
plane of conservation of misery based upon its location in a network.
For example in the core of the network, a cache could be used to move
along the line from C to B without introducing any per-flow
signaling. Another possibility, as suggested above, is to use
premature X2U to move along the line from C to A in an access network
if there is an abundance of bandwidth in the backbone.
7. Control plane
Ooms, et al. Expires April 2002 [Page 12]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
Unlike traditional multicast schemes, Xcast does not specify a
"control plane". There is no IGMP, and as mentioned above, there are
no intradomain or interdomain multicast routing protocols. With
Xcast, the means by which multicast sessions are defined is an
application level issue and applications are not confined to the
model in which hosts use IGMP to join a multicast session. For
example:
- some applications might want to use an IGMP-like receiver-join
model.
- other applications might want to use a model in which a user places
a call to the party or parties that he or she wants to talk to
(similar to the way that one puts together a conference call today
using the button's on one's telephone).
- one might define a session based on the cells that are close to a
moving device in order to provide for a "smooth handoff" between
cells when the moving device crosses cell boundaries.
- in some applications the members of the session might be specified
as arguments on a command line.
- one might define an application that uses GPS to send video from a
bank robbery to the 3 police cars that are closest to the bank being
robbed.
Thus, the application developer is not limited to the receiver-
initiated joins of the IGMP model. There will be multiple ways in
which an Xcast sender determines the addresses of the members of the
channel.
For the purpose of establishing voice and multimedia conferences over
IP networks, several control planes have already been defined,
including SIP [2543] and H.323[H323].
7.1. SIP
In SIP, a host takes the initiative to set up a session. With the
assistance of a SIP server a session is created. The session state
is kept in the hosts. Data delivery can be achieved by several
mechanisms: meshed unicast, bridged or multicast. Note that for the
establishment of multicast delivery, a multicast protocol and
communication with Multicast Address Allocation Servers (MAAS) are
still required.
In "meshed unicast" or "multi-unicasting", the application keeps
track of the participants' unicast addresses and sends a unicast to
Ooms, et al. Expires April 2002 [Page 13]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
each of those addresses. For reasons described in section 3, multi-
unicasting rather than multicast is the prevalent solution in use
today. It's a simple matter to replace multi-unicast code with Xcast
code. All that the developer has to do is replace a loop that sends a
unicast to each of the participants by a single "xcast_send" that
sends the data to the participants. Thus it's easy to incorporate
Xcast into real conferencing applications.
Both Xcast and SIP address super-sparse multicast sessions. It turns
out that Xcast (a very flexible data plane mechanism) can be easily
integrated with SIP (a very flexible control plane protocol). When
an application decides to use Xcast forwarding it does not affect its
interface to the SIP agent: it can use the same SIP messages as it
would for multi-unicasting.
7.2 Receiver Initiated Join model
In the previous section, it was discussed how to establish an Xcast
session among well known participants of a multi-party conference. In
some cases, it is useful for participants to be able to join a
session without being invited. For example, the chairman of a video
chat may want to leave the door of their meeting open for newcomers.
The receiver-initiated join model can be implemented, if desired, by
introducing a server that hosts can talk to to join a conference.
8. Optional information
8.1. List of ports
Although an extension to SIP could be arranged such that all
participants in a session use the same transport (UDP) port number,
in the general case it is possible for each participant to listen on
a different port number. To cover this case, the Xcast packet
optionally contains a list of port numbers.
If the list of port numbers is present, the destination port number
in the transport layer header will be set to zero. On X2U the
destination port number in the transport layer header will be set to
the port number corresponding to the destination of the unicast
packet.
8.2. List of DSCPs
The Xcast packet could (optionally) also contain a list of DiffServ
CodePoints (DSCPs). While traditional multicast protocols have to
create separate groups for each service class, Xcast incorporates the
possibility of having receivers with different service requirements
Ooms, et al. Expires April 2002 [Page 14]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
within one channel.
The DSCP in the IP header will be set to the most demanding DSCP of
the list of DSCPs. This DSCP in the IP header will determine e.g.
the scheduler to use.
If two destinations, with the same next-hop, have 'non-mergable'
DSCPs, two Xcast packets will be created. 'Non-mergable' meaning that
one can not say that one is more or less stringent than the other.
8.3. Channel Identifier
Optionally a sender can decide to add an extra number in the Xcast
header: the Channel Identifier. If the source does not want to use
this option it MUST set the Channel Identifier to zero. If the
Channel Identifier is non-zero the pair (Source Address, Channel
Identifier) MUST uniquely identify the channel (note that this is
similar to the (S, G) pair in SSM). This document does not assign
any other semantics to the Channel Identifier besides the one above.
This Channel Identifier could be useful for several purposes:
1) An identifier of the channel in error, flow control, etc. messages
2) A key to a caching table [OOMS].
3) It gives an extra de-multiplexing possibility (beside the port-
number)
4) ...
9. Encoding
9.1. General
The source address field of the IP header contains the address of the
Xcast sender. The destination address field carries the All-Xcast-
Routers address (to be assigned link-local multicast address), this
is to have a fixed value. Every Xcast router joins this multicast
group. The reasons for putting a fixed number in the destination
field are:
1) The destination address field is part of the IP pseudo header and
the latter is covered by transport layer checksums (e.g. UDP
checksum). So the fixed value avoids a (delta) recalculation of the
checksum.
2) The IPsec AH covers the IP header destination address hence
Ooms, et al. Expires April 2002 [Page 15]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
preventing any modification to that field. Also, both AH and ESP
payloads cover the whole UDP packet (via authentication and/or
encryption). The UDP checksum cannot therefore be updated if the IP
header destination address were to change.
3) In Xcast for IPv6 the Routing Extension shall be used, this header
extension is only checked by a router if the packet is destined to
this router. This is achieved by making all Xcast routers part of
the All_Xcast_Routers group.
4) Normally Xcast packets are only visible to Xcast routers.
However, if a non-Xcast router receives an Xcast packet by accident
(or by criminal intent), it will not send ICMP errors since the Xcast
packet carries a multicast address in the destination address field
([1812]).
Note that some benefits only hold when the multicast address stays in
the destination field until it reaches the end-node (thus not
combinable with X2U).
9.2. IPv4
[AGUI] and [1770] proposed (for a slightly different purpose) to
carry multiple destinations in the IPv4 option. But because of the
limited flexibility (limited size of the header), Xcast will follow
another approach. The list of destinations will be encoded in a
separate header. The Xcast header for IPv4 (in short Xcast4) is
carried between the IPv4 header and the transport layer header.
[IPv4 header | Xcast4 | transport header | payload ]
Note also that since the Xcast header is added to the data portion of
the packet, if the sender wishes to avoid IP fragmentation, it must
take the size of the Xcast header into account.
9.2.1. IPv4 header
The Xcast4 header is carried on top of an IP header. The IP header
will carry the protocol number PROTO_Xcast. The source address field
contains the address of the Xcast sender. The destination address
field carries the All_Xcast_Routers address.
9.2.2. Xcast4 header
The Xcast4 header is depicted in Figure 4. It is composed of two
parts: a fixed part (first 12 octets) and two variable length parts
that are specified by the fixed part.
Ooms, et al. Expires April 2002 [Page 16]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|VERSION|A|X|D|P|R| NBR_OF_DEST | CHECKSUM |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| CHANNEL IDENTIFIER |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| PROT ID | LENGTH | RESV |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| List of Addresses and DSCPs |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| List of Port Numbers (optional) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 4
VERSION = Xcast version number. This document describes version 1.
A = Anonymity bit: if this bit is set the destination addresses for
which the corresponding bit in the bitmap is zero must be overwritten
by zero.
X = Xcast bit: if this bit is set a router must not reduce the Xcast
packet to unicast packet(s), i.e. the packet MUST stay an Xcast
packet end-to-end. This bit can be useful when IPsec is applied.
D = DSCP bit: if this bit is set the packet will contain a DS-byte
for each destination.
P = Port bit: if this bit is set the packet will contain a port
number for each destination.
NBR_OF_DEST = the number of destinations.
CHECKSUM = A checksum on the Xcast header only. This is verified and
recomputed at each point that the Xcast header is processed. The
checksum field is the 16 bit one's complement of the one's complement
sum of all the bytes in the header. For purposes of computing the
checksum, the value of the checksum field is zero. It is not clear
yet whether a checksum is needed (ffs). If only one destination is
wrong it can still be useful to forward the packet to N-1 correct
destinations and 1 incorrect destination.
CHANNEL IDENTIFIER = 4 octets Channel Identifier (see section 8.3).
Since it is located within the first 8 bytes of the header, it will
be returned in ICMP messages.
PROT ID = specifies the protocol of the following header.
Ooms, et al. Expires April 2002 [Page 17]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
LENGTH = length of the Xcast header in 4-octet words. This field
puts an upper boundary to the number of destinations. This value is
also determined by the NBR_OF_DEST field and the D and P bits.
RESV = R = Reserved. It must be zero on transmission and must be
ignored on receipt.
The first variable part is the 'List of Addresses and DSCPs', the
second variable part is the 'List of Port Numbers'. Both are 4-octet
aligned. The second variable part is only present if the P-bit is
set.
Figure 5 gives an example of the variable part for the case that the
P-bit is set and the D-bit is cleared (in this example N is odd):
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BITMAP |
~ ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Destination 1 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
~ ... ~
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Destination N |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Port 1 | Port 2 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
~ ... ~
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Port N | padding |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 5
BITMAP = every destination has a corresponding bit in the bitmap to
indicate whether the destination is still valid on this branch of the
tree. The first bit corresponds to the first destination in the
list. This field is 4-octet aligned (e.g. for 49 destinations there
will be a 64-bit bitmap). If Xcast is applied in combination with
IPsec, the bitmap - since it can change on route - has to be moved to
a new to be defined IPv4 option.
List of Destinations. Each address size is four octets.
List of Port Numbers. List of two octet destination port number(s),
where each port corresponds in placement to the preceding Destination
Ooms, et al. Expires April 2002 [Page 18]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
Address.
9.3. IPv6
The Xcast6 header encoding is similar to IPv4, except that Xcast
information is stored in IPv6 extension headers.
[IPv6 header | Xcast6 | transport header | payload ]
9.3.1. IPv6 header
The IPv6 header will carry the NextHeader value 'Routing Extension'.
The source address field contains the address of the Xcast sender.
The destination address field carries the All_Xcast_Routers address.
9.3.2. Xcast6 header
The Xcast6 header is also composed of a fixed and two variable parts.
The fixed and the first variable part is carried in a Routing
Extension. The second variable part is carried in a Destination
Extension.
9.3.2.1. Routing Extension header
The P-bit of Xcast4 is not present because it is implicit by the
presence or absence of the Destination Extension (Figure 6).
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next Header | HdrExtLen |RouteType=Xcast| 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|VERSION|A|X|D| R | NBR_OF_DEST | CHECKSUM |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| CHANNEL IDENTIFIER |
~ ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| List of Addresses and DSCPs |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 6
HdrExtLen = The header length is expressed in 8-octets, thus a
maximum of 127 destinations can be listed (this is why NBR_OF_DEST is
7-bit).
RouteType = Xcast should be assigned by IANA.
Ooms, et al. Expires April 2002 [Page 19]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
The fourth octet is set to 0.
R = Reserved.
CHANNEL IDENTIFIER = 16 octets Channel Identifier (see section 8.3).
The other fields are defined in section 9.2.2.
The 'List of Addresses and DSCPs' is 8-octet aligned. The size of
the bitmap is determined by the number of destinations and is a
multiple of 64 bits.
9.3.2.2. Destination Extension header
Optionally the Destination Extension (Figure 7) is present to specify
the list of Port Numbers. The destination header is only evaluated
by the destination node.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next Header | HdrExtLen |Opt Type=Ports | Opt Data Len |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| List of Port Numbers |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 7
Option Type for Ports should be assigned by IANA. The first three
bits MUST be 010 to indicate that the packet must be discarded if the
option is unknown and that the option can not be changed en-route.
The number of Ports MUST be equal to the number of destinations
specified in the Routing header.
10. Impact on Upper Layer Protocols
Some fields in the Xcast header(s) can be modified as the packet
travels along its delivery path. This has an impact on:
10.1. Checksum calculation in transport layer headers
In transport layer headers, the target of the checksum calculation
includes the IP pseudo header, transport header and payload (IPv6
header extensions are not a target).
The transformation of an Xcast packet to a normal unicast packet -
Ooms, et al. Expires April 2002 [Page 20]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
(premature) X2U - replaces the multicast address in the IP header
destination field by the address of a final destination. If the
Xcast header contains a Port List, the port number in the transport
layer (which should be zero) also needs to be replaced by the port
number corresponding to the destination. This requires a
recalculation of these checksums. Note that this does not require a
complete recalculation of the checksum, only a delta calculation,
e.g. for IPv4:
Checksum' = ~ (~Checksum + ~daH + ~daL + daH' + daL' + ~dp + dp')
In which "'" indicates the new values, "da" the destination address,
"dp" the destination port and "H" and "L" respectively the higher and
lower 16 bit.
10.2. IPsec
This is described in [PARI].
11. Gradual Deployment
11.1. Tunneling
One way to deploy Xcast in a network that has routers that have no
knowledge of Xcast is to setup "tunnels" between Xcast peers (MBone
approach). This enables the creation of a virtual network layered on
top of an existing network [2003]. The Xcast routers exchange and
maintain Xcast routing information via any standard unicast routing
protocol (e.g. RIP, OSPF, ISIS). The Xcast routing table that is
created is simply a standard unicast routing table that contains the
destinations that have Xcast connectivity, along with their
corresponding Xcast next hops. In this way, packets may be forwarded
hop-by-hop to other Xcast routers, or may be "tunneled" through non-
Xcast routers in the network.
For example, suppose that A is trying to get packets distributed to
B, C & D in Figure 8 below, where "X" routers are Xcast-capable, and
"R" routers are not. Figure 9 shows the routing tables created via
the Xcast tunnels:
Ooms, et al. Expires April 2002 [Page 21]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
R4 ---- B
/
/
A ----- X1 ---- R2 ---- X3 R8 ---- C
\ /
\ /
R5 ---- R6 ---- X7
\
\
R9 ---- D
Figure 8
Router X1 establishes a tunnel to Xcast peer X3. Router X3
establishes a tunnel to Xcast peers X1 and X7. Router X7 establishes
a tunnel to Xcast peer X3.
X1 routing table: X3 routing table: X7 routing table:
Dest | NextHop Dest | NextHop Dest | NextHop
------+---------- ------+--------- ------+---------
B | X3 A | X1 A | X3
C | X3 C | X7 B | X3
D | X3 D | X7
Figure 9
The source A will send an Xcast packet to its default Xcast router,
X1, that includes the list of destinations for the packet. The packet
on the link between X1 and X3 is depicted in Figure 10:
Ooms, et al. Expires April 2002 [Page 22]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
+----------+
| payload |
+----------+
| UDP |
+----------+
| Xcast |
| B,C,D |
| prot=UDP |
+----------+
| inner IP |
| src=A |
|dst=All_X_|
|prot=Xcast|
+----------+
| outer IP |
| src=X1 |
| dst=X3 |
| prot=IP |
+----------+
Figure 10
When X3 receives this packet, it processes it as follows:
- Perform a route table lookup in the Xcast routing table to
determine the Xcast next hop for each of the destinations listed in
the packet.
- If no Xcast next hop is found, replicate the packet and send a
standard unicast to the destination.
- For those destinations for which an Xcast next hop is found,
partition the destinations based on their next hops.
- Replicate the packet so that there's one copy of the packet for
each of the Xcast next hops found in the previous steps.
- Modify the list of destinations in each of the copies so that the
list in the copy for a given next hop includes just the destinations
that ought to be routed through that next hop.
- Send the modified copies of the packet on to the next hops.
- Optimization: If there is only one destination for a particular
Xcast next hop, send the packet as a standard unicast packet to the
destination, since there is no advantage to forwarding it as an Xcast
packet.
So, in the example above, X1 will send a single packet on to X3 with
Ooms, et al. Expires April 2002 [Page 23]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
a destination list of < B C D >. This packet will be received by R2
as a unicast packet with destination X3, and R2 will forward it on,
having no knowledge of Xcast. When X3 receives the packet, it will,
by the algorithm above, send one copy of the packet to destination <
B > as an ordinary unicast packet, and 1 copy of the packet to X7
with a destination list of < C D >. R4, R5, and R6 will behave as
standard routers with no knowledge of Xcast. When X7 receives the
packet, it will parse the packet and transmit ordinary unicast
packets addressed to < C > and < D > respectively.
11.2. Premature X2U
If a router discovers that its downstream neighbor is not Xcast
capable, it can perform a Premature X2U, i.e. send a unicast packet
for each destination in the Xcast header which has this neighbor as a
next hop. Thus duplication is done before the Xcast packet reached
its actual branching point.
A mechanism (protocol/protocol extension) to discover the Xcast
capability of a neighbor is ffs. Among others, one could think of an
extension to a routing protocol to advertise Xcast capabilities or
one could send periodic 'Xcast pings' to its neighbors (send an Xcast
packet that contains its own address as a destination and check
whether the packet returns).
11.3. Semi-permeable tunneling (IPv6 only)
This is an optimization of tunneling in the sense that it does not
require (manual) configuration of tunnels. It is enabled by adding a
Hop-by-Hop Xcast6 header. An IPv6 packet can initiate/trigger
additional processing in the on-route routers by using the IPv6 Hop-
by-hop option.
The type of the Xcast6 Hop-by-hop option has a prefix '00' so that
routers that cannot recognize Xcast6 can treat the Xcast6 datagram as
a normal IPv6 datagram and forward toward the destination in the IPv6
header.
Packets will be delivered to all members if at least all
participating hosts are upgraded.
When the source A sends an Xcast packet via semi-permeable tunneling
to destinations B, C and D it will create the packet of Figure 11.
One of the final destinations will be put in the destination address
field of the outer IP header.
Ooms, et al. Expires April 2002 [Page 24]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
+----------+
| payload |
+----------+
| UDP |
+----------+
| Xcast |
| |
+----------+
| inner IP |
| src=A |
|dst=All_X_|
|prot=Xcast|
+----------+
| Xcast |
|SP-tunnel |
|Hop-by-hop|
+----------+
| outer IP |
| src=A |
| dst=B |
| prot=IP |
+----------+
Figure 11
Semi-permeable tunneling is a special tunneling technology that
permits intermediate Xcast routers on a tunnel to check the
destinations and branch if destinations have a different next hop.
Note that with the introduction of an Xcast IPv4 option, this
technique could also be applied in IPv4 networks.
11.4. Special case: deployment without network support
A special method of deploying Xcast is possible by upgrading only the
hosts. By applying tunneling (see section 11.1 and 11.3) with one of
the final destinations as tunnel endpoint, the Xcast packet will be
delivered to all destinations when all the hosts are Xcast aware.
Both normal and semi-permeable tunneling can be used.
If host B receives this packet, in the above example, it will notice
the other destinations in the Xcast header. B will create a new
Xcast packet and will send it to one of the remaining destinations.
In the case of Xcast6 and semi-permeable tunneling, Xcast routers can
be introduced in the network without the need of configuring tunnels.
The disadvantages of this method are that:
Ooms, et al. Expires April 2002 [Page 25]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
- all hosts in the session need to be upgraded.
- non-optimal routing.
- anonymity issue: hosts can know the identity of other parties in
the session (which is not a big issue in conferencing, but maybe for
some other application?).
- host has to perform network functions and needs an upstream link
which has the same bandwidth as its downstream link.
12. (Socket) API
In the most simple use of Xcast, the final destinations of an Xcast
packet receive an ordinary unicast UDP packet. This means that hosts
can receive an Xcast packet with a standard, unmodified TCP/IP stack.
Hosts can also transmit Xcast packets with a standard TCP/IP stack
with a small Xcast library that sends Xcast packets on a raw socket.
This has been used to implement Xcast based applications on both Unix
and Windows platforms without any kernel changes.
Another possibility is to modify the sockets interface slightly. For
example, one might add an "xcast_sendto" function that works like
"sendto" but that uses a list of destination addresses in place of
the single address that "sendto" uses.
13. Security Considerations
See [PARI].
References
[1075] D. Waitzman, C. Partridge, S.E. Deering, Distance Vector Multi-
cast Routing Protocol, RFC 1075, November 1988.
[1770] C. Graff, "IPv4 Option for Sender Directed Multi-Destination
Delivery", RFC1770, March 1995.
[1812] F. Baker, "Requirements for IP Version 4 Routers", RFC1812, June
1995.
[2003] Perkins, C., "IP Encapsulation within IP", RFC 2003, October
1996.
Ooms, et al. Expires April 2002 [Page 26]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
[2201] A. Ballardie, Core Based Trees (CBT) Multicast Routing Architec-
ture, RFC 2201, Sept. 1997.
[2236] W. Fenner, Internet Group Management Protocol, Version 2, RFC
2236, Nov. 1997.
[2401] S. Kent, R. Atkinson, "Security Architecture for the Internet
Protocol", RFC2401, November 1998.
[2460] S. Deering, R. Hinden. Internet Protocol, Version 6 (IPv6),
RFC2460, December 1998.
[2543] M. Handley, H. Schulzrinne, E. Schooler, J, Rosenberg, "SIP:
Session Initiation Protocol", RFC2543, March 1999.
[2902] S. Deering, S. Hares, C. Perkins, R. Perlman, "Overview of the
1998 IAB Routing Workshop", RFC2902, August 2000.
[AGUI] L. Aguilar, "Datagram Routing for Internet Multicasting",
Sigcomm84, March 1984.
[BOIV] R. Boivie, N. Feldman, "Small Group Multicast", draft-boivie-
sgm-01.txt, July 2000.
[DEER] S. Deering, "Multicast Routing in a datagram internetwork", PhD
thesis, December 1991.
[DEE2] S. Deering, D. Estrin, D. Farinacci, V. Jacobson, C. Liu, and L.
Wei. The Pim Architecture for Wide-area Multicast Routing, ACM
Transactions on Networks, April 1996
[DOOR] B. Van Doorselaer, D. Ooms, "SIP for the establishment of
xcast-based multiparty conferences", draft-van-doorselaer-sip-
xcast-00.txt, July 2000.
[FARI] D. Farinacci, "Multicast Source Discovery Protocol", draft-
farinacci-msdp-00.txt, June 1998.
[H323] ITU-T Recommendation H.323 (2000), Packet-Based Multimedia Com-
munications Systems.
[HOLB] H. Holbrook, B. Cain, "Source-Specific Multicast for IP",
draft-holbrook-ssm-00.txt, March 2000.
[IMAI] Y. Imai, Multiple Destination option on IPv6 (MDO6), <draft-
imai-mdo6-02.txt>, September 2000
[MBONE] Frequently Asked Questions (FAQ) on the Multicast Backbone
Ooms, et al. Expires April 2002 [Page 27]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
(MBONE), ftp://venera.isi.edu/mbone/faq.txt
[OOMS] D. Ooms, W. Livens, Connectionless Multicast, <draft-ooms-cl-
multicast-02.txt>, April 2000
[PARI] O. Paridaens, D. Ooms, Security Framework for Explicit Multi-
cast, draft-paridaens-xcast-sec-framework-01.txt, November 2000.
[PERL] R. Perlman, "Simple Multicast: A design for Simple, Low-overhead
Multicast", draft-perlman-simple-multicast-02.txt, February
1999.
[RMT] Reliable Multicast Transport Working Group web site,
http://www.ietf.org/html.charters/rmt-charter.html, June 15,
1999
[SOLA] M. Sola, M. Ohta, T. Maeno. Scalability of Internet Multicast
Protocols, INET'98,
http://www.isoc.org/inet98/proceedings/6d/6d_3.htm
Authors Addresses
Rick Boivie
IBM T. J. Watson Research Center
30 Saw Mill River Rd.
Hawthorne, NY 10532
Phone: 914-784-3251
Email: rhboivie@us.ibm.com
Nancy Feldman
IBM T. J. Watson Research Center
30 Saw Mill River Rd.
Hawthorne, NY 10532
Phone: 914-784-3254
Email: nkf@us.ibm.com
Yuji Imai
Fujitsu LABORATORIES Ltd.
1-1, Kamikodanaka 4-Chome, Nakahara-ku, Kawasaki 211-8588, Japan
Phone : +81-44-754-2628
Fax : +81-44-754-2793
E-mail: kimai@flab.fujitsu.co.jp
Ooms, et al. Expires April 2002 [Page 28]
Internet Draft draft-ooms-xcast-basic-spec-02.txt October 2001
Wim Livens
Colt Telecom
Zweefvliegtuigstraat 10, 1130 Brussel, Belgium.
Phone : 32 2 7901705
Fax : 32 2 7901711
E-mail: WLivens@colt-telecom.be
Dirk Ooms
Alcatel Network Strategy Group
Fr. Wellesplein 1, 2018 Antwerpen, Belgium.
Phone : 32 3 2404732
E-mail: Dirk.Ooms@alcatel.be
Olivier Paridaens
Alcatel Network Strategy Group
Fr. Wellesplein 1, 2018 Antwerpen, Belgium.
Phone : 32 3 2409320
E-mail: Olivier.Paridaens@alcatel.be
Ooms, et al. Expires April 2002 [Page 29]