[Chapter-delegates] MMM # 15 with Jose Morales-Barroso PhD + TRANSCRIPTION

Tue Mar 21 17:45:49 PST 2006

See below, the TRANSCRIPTION of our MMM meeting today.

For all those who couldn’t attend we’ll held another edition of this session
probably next week (after the IETF).

I strongly recommend to review the presentation he prepared for today’s
lecture  <http://www.lmdata.es/uets/uets-isoc.pdf>
http://www.lmdata.es/uets/uets-isoc.pdf

and also follow the final discussion among the lecturer (Jose Morales below
as “etherforum”) Edmon Chung (Hong Kong Chapter) and Hans Peter Dittler
(German Chapter).

         andreu

Invited Guest: Jose Morales Barroso, PhD

edmon_chung has joined the conference.
David McAuley has joined the conference.
Jim has joined the conference.
Hans Peter Dittler has joined the conference.
etherforum has joined the conference.
David McAuley: Hello all from Dallas
ruudisoc: Hello from Ghent - Belgium
Andreu: Hello from cloudy Stanford (California)
edmon_chung: Hi from Dallas too
Hans Peter Dittler: Hello also from Dallas (IETF)

  _____  

Andreu: Today we've a great speaker with us
Andreu: Let me introduce Jose Morales PhD
Andreu: He is an Spanish engineer 
Andreu: involved in the technologies of data communications since 1976,
Andreu: During his "free time" he 
Andreu: collaborates with the Development of the Ethernet IEEE 802.3
Andreu: Working Groups
Andreu: And has a very radical and original idea
Andreu: on how to scale the internet.
Andreu: He has prepared a nice presentation
Andreu: that I sent to Chapter delegates an hour ago
Andreu: but didnt received yet...
Andreu: Please
Andreu: download it from here 
David McAuley: I saw the note to chapters and did download it
Vittorio Bertola has joined the conference.
Vittorio Bertola: good night, everyone 
ruudisoc: the slides are impressive
Andreu: Vittorio you missed this line 
Andreu:  <http://www.lmdata.es/uets/uets-isoc.pdf>
<http://www.lmdata.es/uets/uets-isoc.pdf>
http://www.lmdata.es/uets/uets-isoc.pdf
Andreu: This is the (last) correct one
Andreu: Let's leave a couple of minutes to download
Andreu: and please keep the presentation open
etherforum: Hello from Madrid
Andreu: NOTE: you can add Jose's IM ID to your accounts
Andreu: he's "etherforum"
etherforum: There is a little raining
Andreu: So everybody has the presentation on ?
David McAuley: yes
edmon_chung: got it
Andreu: Jose, when you want we can start.
etherforum: It is great to be with you today (night in Spain)
Hans Peter Dittler: have the paper ready and skimmed through it...
etherforum: I will try to to present the new UETS/EFR Reference Model to
you.
Andreu: Jose is a member of our ISOC-ES community
Andreu: and his ideas got the attention of IEEE
Andreu: where he has published recently a paper
Andreu: We thought that most of you would be interested
David McAuley: That is great - thanks Jose for doing this today
Andreu: on hearing this ideas. 
Andreu: Which are really different of what we normally see.
etherforum: To understand it, I will try to explain the "genesis" to you.
etherforum: Let me remember some history:
etherforum: 1973 was the year of the "Big Bang": Ethernet & TCP/IP was born.
etherforum: In the "Packetozoic Era" (<> 1976- 1980), the CPU speed (4 MHz
in the INFONET nodes, the RCCs) was MUCH higher than the WAN speed (9.600
bps Modems)
etherforum: Vint Cerf developed the architecture of Internet from Louis
Pouzin’s "catenet" idea, as described in the IEN48 document: it was
necessary of creating a "confederation of cooperating nets".
etherforum: The TCP was developed "to interconnect these networks in such a
way
 that would allow many such networks to interwork and the computers on
each of them to interwork".
etherforum: This permits the existence of a unique logic net upon the many
physical networks.
etherforum: Vint Cerf and Bob Kahn decided to pass the control of the
communication to the host (TCP), leaving then the nodes of the net
(gateway/routes IP) working only in communication tasks in a network of
datagrams.
etherforum: In 1990, many countries tried to kill TCP/IP: remember, for
example GOSIP "Federal Information Processing Standard", but IP was the
winner, because its simplicity. 
etherforum: But today, IP does not scale, because it is designed to be
software. As a matter of fact, the technical people have developed MPLS to
do switching instead of routing. But MPLS makes label swapping, and it
depends of the speed of the memories

etherforum: Now, we are in the dawn of a "New Era": the speed of the WAN (10
Giga Ethernet, and 100 Giga Ethernet in the near future) is higher than the
CPU (4 GHZ) and much more than the memory speed, which does not scale.
etherforum: In March, 2004, I proposed the Ethernet Power Management for
energy saving, during my participation in the final balloting voting phase
of the IEEE 802.3ah EFM project.
etherforum: You can see my proposal in:
<http://www.lmdata.es/uets/uets-epm.pdf>
http://www.lmdata.es/uets/uets-epm.pdf

etherforum: As a result of a study and deeper reflection I have developed
the UETS system. 
etherforum: While I was preparing my presentation about how to increase the
functions of Ethernet, I studied the documents of "Backplane Ethernet", and
discovered that the best solution to communicate all the devices in a
multi-processor system was to use the LLC 802.2 protocols.
etherforum: Then, I sent a presentation to use LLC:
 <http://www.ieee802.org/3/cm_study/email/pdf00000.pdf>
http://www.ieee802.org/3/cm_study/email/pdf00000.pdf

etherforum: In 1987 I was member of the "high level design group" of Tesys
B, the second generation of nodes for the X.25 Spanish network. Some of the
people in the team had developed the world’s first public packet network,
the Telefonica RSAN.
etherforum: The bus/backplane architecture of the first generation (Tesys A)
did not scale. Therefore, we decided to use a topology based in a MATRIX
switch for the interconnection of the processors.
etherforum: One night, reading the reflector, I applied the the MATRIX to do
a switch in order to solve the scaling of the multiprocessor systems based
in Ethernet, using directly the local (bit U/L = 1) MAC address as label, in
order to be able to use the PHYSICAL address as NETWORK address.
etherforum: It is very important the LOCAL address, because all the current
internetworking equipment use the UNIVERSAL addressing.
etherforum: I was completely shocked when I discovered that the network
could scale to more than 70 trillion terminals! 
etherforum: Although I have experience in the design of switching equipment,
that used this type of techniques, I find surprised the fact that nobody had
reached this idea before.
etherforum: With this huge addressing, the switching technique, that I call
EFR (Ethernet Fabric Routing) can be used not only in the multi-processor
environments, but also in LAN/MAN/WAN.
etherforum: For routing the Ethernet frames in the core it will be possible
to use the techniques of ATM/MPLS with the MAC address as a label, but
without the swapping that limits the speed due to the limitations of the
memory (DRAM or SRAM).
etherforum: The possibility of telecommunication connection with Ethernet
opens a radical new approach to the digital revolution... 
etherforum: The idea is very simple: TCP/IP was developed to
"internetworking" because in 1976 the networks doesn’t scale. But now, UETS
permits a very big network.
etherforum: If you use the MAC-in-MAC (802.1 standard) or the classical SNAP
(802 standard) and VLANs, the addressing is:
etherforum: 6 (MAC) + 2 (VLAN) + 6 (MAC2) + 2 (VLAN2) = 16 octets
etherforum: It is like IPv6 addressing, but routerless!!
etherforum: In UETS/EFR, the transport of IP over Ethernet is perfect, but
in order to improve the Internet capabilities, I use a double stack of
protocols: the TCP/IP for IP internetworking and LLC/ETH for the
communication inside the Ethernet domain.
etherforum: The sockets, that are intact in the application part, can use
for the transport the LLC2 protocol instead of the TCP circuits, the LLC1 or
LLC2 instead of the UDP datagram and it is directly over Ethernet frames
with its address of six octets.
etherforum: Even considering only one Ethernet domain, there is a paradox:
the physical network Ethernet has much more addresses (6 octets) than the
logical network IP (4 octets).
etherforum: If you replace the IP (datagrams) protocol with Ethernet
(datagrams), TCP with LLC type 2, and UDP with LLC type 1, the Internet
applications (sockets) will operate over Ethernet networks. 
etherforum: Moreover, this improves in an extraordinary manner the security
of the net, given that the use of the addresses of the net for the switching
makes impossible to fake them. In UETS/EFR model, the PHYSICAL and LOGICAL
networks are the same.
etherforum: This radical new approach, maintain untouched the original idea
and "The Spirit of Internet".
etherforum: It does not seem to you interesting that so looked for
convergence of networks was obtained with the oldest protocols: HDLC and
Ethernet? I believe that it is not only possible, but also desirable!!!
etherforum: The UETS/EFR page is available [on-line]: 
 <http://www.lmdata.es/uets.htm> http://www.lmdata.es/uets.htm

etherforum: I would be very happy to supply you any further information that
you may want.
David McAuley: Thank you for this. I would like to pass this information on
to other interested parties - I assume that is ok but wanted to ask 
etherforum: Ok
David McAuley: Thanks you Jose - very interesting and I will include a copy
to Bob metcalfe as well
Andreu: WoW
David McAuley: I methim some years ago at Polaris ventures
Andreu: As always we leave 30 minutes for the lecturer
Andreu: and now we can open the debate 
Andreu: or questions.
Hans Peter Dittler: This was an overwhelming amount of information - but for
me it opens up more questions than answers
etherforum: Thanks, Bob is aware of this proposal, and also Vint Cerf. But
it will be nice if you pass to him this presentation
Andreu: (David, I copied him today inviting him to download the
presentation, but please go ahead).
etherforum: I will try to answer your questions
etherforum: The model is extremely simple
edmon_chung: sounds like a very interesting idea... though i m not sure
about one thing (maybe a stupid question)... is the proposal to utilize MAC
addresses in place of IP addressing scheme?... in that case what happens to
extensibility?... right now, we at least can "arguably" (dont want to go
entirely there) move from v4 to v6, how would the proposal fair wrt future
extensibility... 
Hans Peter Dittler: I can follow you on having this concept working in a
homogenous directly connected network but i can not see the extensions to a
worldwide routing function
etherforum: As I told you before, the ETH domain wil extend to 70 trillion
addresses
etherforum: The UETS system does not use routing, it is pure switching,
without label swapping
Hans Peter Dittler: Each switch will have to hold the complete architecture
of the full network - that seems a little bit huge...
etherforum: No, the addres is the I/F address
etherforum: It works like the old telephone network
etherforum: Please, look at page 20 "CUE implementation: L2 bridge, L3 sw,
IP router, ATM sw, FCH fabric"
etherforum: The model is so simple that is very difficult to believe
edmon_chung: ic
Hans Peter Dittler: yes ok - that is exactly my problem - this switch must
then have as many interfaces as there are nodes in the network??
etherforum: No, look at figure 21. 
etherforum: The MAC address carries the physical ID of the switch
edmon_chung: what happens when machines move around then?... how does it get
switched?...
Hans Peter Dittler: OK - that is the concept of ATM or X.25 - then you
suddenly have to have a hierachical addressing scheme, where your address
gives the hierachical (geografically hierachical) routing or transport or
switching information
etherforum: As you are the administrator, the addresing can be as you prefer
Hans Peter Dittler: Somebody has to assign all the addresses fitting to this
scheme to the end-nodes, my laptop would get a new address depending on the
hot-spot I am connected to - how is that handled?
etherforum: But this is a DATAGRAMS network not a CO network of virtual
circuits
edmon_chung: ah... so its a combo MAC+geographic for the designation of a
terminal/machine?...
etherforum: Yout laptop use the universal address, the TRUE is in charge of
convert this addres to the switching
Andreu: (sorry, what's the "TRUE" ?) 
etherforum: The architecture of UETS/EFR and its rtealtion with the current
switches is describes in page 19
Andreu: Ops I see, thanks.
Hans Peter Dittler: oops - then You were talking about a new infrastructure
network and not about a replacement for the Internet as seen by the
end-user??
etherforum: There are 3 basic elements: terminals (TUE) access device (TRUE)
and switches (CUE)
etherforum: No, it is not a new infrastructure, because it is 100% Ethernet
Hans Peter Dittler: Why should Ethernet not be an infrastructure?
Andreu: Well Ethernet is already "deployed" so he means
etherforum: I call infrastructure the physical network, and ethernet is also
half link layer, the other half, according IEEE 802, should be IEEE 802.2
LLC
Andreu: its not a "new infrastruture"
etherforum: Ethernet is deployed in campus/local, but now it is been
deployed in acces and MAN
etherforum: There are sever problems of scalability with the current
techniques
etherforum: UETS/EFR permits the evolution of Internet without migration
Andreu: Less "intrusive" than IPv6 then ?
edmon_chung: just trying to visualize ... so in your proposal, basically it
is still packet switched but the IP envelope will be gone, instead, the
envelope itself utilizes the physical addresses, so, e.g. lets say there
will still be DNS, then a domain will be mapped to one of the universal
address?
etherforum: As you know, IPv6 and IPv4 are different protocols, and both
offer the network services at layer 4
etherforum: In UETS ther is the Ethernet DNS (EDNS)
etherforum: It is exactly the Internet model, but only in one net. You can
see in figure 14 that there are redundancies using IP over ETH
Hans Peter Dittler: now you have lost me - you say there is no migration
necessary but you replace DNS by something else - that is for me a very
radical migration...
etherforum: In UETS you will have some services, like file servers, video on
demand, etc, and at the same time maintain the IP connections.
edmon_chung: so you mean you are suggesting it could be TCP/IP/UETS?...
etherforum: You use LLC/ETH only to some applications. The IP applications
use the classic DNS, of course.
etherforum: Yes, the stack is double, as described in page 15
edmon_chung: ah... ic.. so existing apps can still run over an IP layer, but
you are proposing an option to strip away the IP envelope
etherforum: There is no problem at all to use two stacks
etherforum: If you use the LLC, the socket will be open at layer 2.
Jim: Sorry folks. I've been lurking but I need to go. It's been GREAT!
THANKS!
Jim has left the conference.

Hans Peter Dittler: With parallel stacks you go down just the same ratholes
as IPv6 regarding necessity to change applications and services
etherforum: There is no need to change apllications or services, some
applications open the TCP/UDP, and other opens LLC
Andreu: It's just a "software" (stack) change, then ?
Andreu: To replace the actual TCP/IP stacks by dual
Andreu: stacks ?
etherforum: Yes, ther is only a piece of software
etherforum: Actually, I use a similar technique in my company
Hans Peter Dittler: That is exactly the same story I tell all my customers
when they change to IPv6 - it is not true! Nearly all applications have
embedded addresses somewhere and these have to bee changed
etherforum: The local services are over IPX (Novell) and the connetion to
Internet with TCP/IP
etherforum: With UETS, you dont need to change the current IPv4
applications, because it will work over the same stack
edmon_chung: have a different question ... not that familiar with MAC
addresses... can i get a virtual MAC address?... if not, what happens if i
want 2 UETS address for 1 machine?...
David McAuley: Rudi - you will archive this is that right
ruudisoc: sorry just came back was all the time on the phone
etherforum: There is no problem at all: only connect the terminal to 2 I/F
ruudisoc: yes of course I'll fil eit
David McAuley: Thanks
Andreu: OK
Andreu: This is becoming very interesting
Andreu: but we're runing out of time
ruudisoc: but i'm not sure I have it all in my chat window here 
Andreu: I think we should do a second edition 
ruudisoc: Andreu can you also copy and paste it in an email to me
David McAuley: OK I will save copy also and can mail tomorrow if neeed but
am on the run today
Andreu: of this session, next week
Andreu: (sure, Rudi)
Hans Peter Dittler: It was very interesting - I will lean back and wait if
this new idea can come up and exist and replace others like ATM or MPLS for
transport - thanks a lot
David McAuley: yes thank you Jose
Hans Peter Dittler: see you next time - bye from Dallas
etherforum: Thanks everybody. Bye. 
Hans Peter Dittler has left the conference.

ruudisoc: very interesting stuff, need more time to explore possibilities
etherforum: Remember the UETS page:  <http://www.lmdata.es/uets.htm>
http://www.lmdata.es/uets.htm
ruudisoc: thanks to you Jose
Andreu: Jose 
edmon_chung: very interesrting 
Andreu: maybe you can give them 
ruudisoc: perhaps we can continue the debate in the forum
edmon_chung: bye
Andreu: your email
Andreu: or contact data.
Andreu: But probably is all in the link
Andreu: right ?
etherforum: my email: xxxxxxxxxxxxxxxxxx
David McAuley: Thanks and we will, if you can, try to arrange another
Andreu: Sure. But next week 
ruudisoc: questions can be posted in the forum aside the script of this
meeting from this night on
Andreu: when IETF is over
David McAuley: I am afraid I must run - thanks everyone and Jose a special
thanks to you
etherforum: It is a pleasure.
Andreu: Thanks to all for attending
Andreu: it has been an incredible session.
Andreu: with a lot of info.
Andreu: ----------------------------------
Andreu: We must follow this
Andreu: UETS proposal
Andreu: See you soon.
Andreu: OVER & OUT

  _____  

De: Andreu Veà [mailto:andreu at veabaro.info] 
Enviado el: martes, 21 de marzo de 2006 11:08
Para: 'chapter-delegates at elists.isoc.org'
CC: 'Junta Directiva de ISOC-ES'; 'Jose Morales Barroso'
Asunto: REMINDER: Meeting 1 hour from now : MMM # 15 with Jose
Morales-Barroso PhD + SEE Slides

Jose has prepared a terrific document to illustrate his presentation for our
meeting today:

Please download it from here     http://www.lmdata.es/uets/uets-isoc.pdf and
have it ready during his lecture.

Have your Messenger on, and I’ll invite you to join. (send me an email if
you need directions).

I blind copy Vint Cerf and Bob Metcalfe, cause I know they’ll probably be
interested in these ideas.

            andreu

 MONOGRAPHIC MONTHLY MEETING:    #   15             MARCH – 2006

------------------------------------8<--------------------------------------
-------------------------------
    Date:  Tuesday,  March the 21st . 
   Time:  20:00 UTC  Look for your city at the newly prepared meeting
planner below . 

    Topic:         “Ethernet Everywhere. 

      The dawn of the Net [and/means] the decline of the Internet ?“

      A very radical an original approach to how to keep developing the
internet. An IP v6 alternative ?

    Guest :      **  Jose Morales-Barroso PhD **  ISOC-ES & IEEE Member.

         PROFILE:  He has a long lasting career in the field of Computer
Networks. In 1976 he worked as a Technical Systems Engineer in INFONET. 

Between 1986 and 1987, he was part of the design team of TESYS B/X.25,
packet switching node developed for Telefónica, 

together with the engineers who developed the RSAN (1971), the world’s first
public packet-switched network. 

In 1988, he founded L&M Data Communications, where he has taught over 900
courses in Data Comms. 

    Status:    Confirmed
----------8<----------------------------------------------------------------
------------------------------- 

HOW TO ATTEND ? 

Just have your Yahoo Messenger, Gaim/Jabber
 ON,  and you'll be invited 5
minutes before.

New-Timers: Send me an email to subscribe you to the meeting  landreu (at)
gmail (dot) com  up to 10 minutes before the event start time.  

Come and join us. This time many people from the IETF are invited too.

Thanks, 

       andreu

Don't miss this unique oportunity to know more about our ISOC colleagues
around the world and collect first hand information from our monthly
"invited guests". 

Take advantage of being able to opine about what else can be done. In the
partnership between ISOC and the ISOC Chapters. Its the opportunity and the
perfect occasion to express your ideas. 

.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://elists.isoc.org/mailman/private/chapter-delegates/attachments/20060321/e9880568/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.emz
Type: application/octet-stream
Size: 345632 bytes
Desc: not available
URL: <https://elists.isoc.org/mailman/private/chapter-delegates/attachments/20060321/e9880568/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image003.gif
Type: image/gif
Size: 7327 bytes
Desc: not available
URL: <https://elists.isoc.org/mailman/private/chapter-delegates/attachments/20060321/e9880568/attachment.gif>