Jxta - An emerging Internet Operationg System?

[Doug Wang]

Question: What exactly is Jxta?
Answer: A set of open peer-to-peer protocols that will enable
the development of new network computer applications.
Jxta should be an emerging Inernet operating system. Jxta seems to be equivalent to Microsoft's Windows for P2P era. What's more, such "Internet" doesnt depend on TCP/IP. It is a more pervasive Internet. Any digital device may be able to plug into it.
Any thoughts?
[ February 14, 2002: Message edited by: Doug Wang ]

[Michael Ernest]

I don't think you want to go so far as to say the Internet doesn't depend on TCP/IP. It does. Ready to hear a concrete counter-example though.

[Doug Wang]

Thanks Michael for correcting me.
I should have used P2P Network instead of Internet. But in the near future, this P2P Network, going alongside with the existing Internet, is to form another pervasive Internet.
Correct me if i'm wrong about the concept.
[ February 14, 2002: Message edited by: Doug Wang ]
[ February 15, 2002: Message edited by: Doug Wang ]

[Latha Kalaga]

I am not sure whether I fully understand the term, "pervasive Internet". Every kind of networking does depend on some means of transport and the most popular protocol is TCP/IP. I have found an article Project JXTA: An Open, Peer-to-Peer Collaboration Platform using Java and XML
which states that JXTA can be implemented independent of a transport protocol (ie., internet, intranet, private-network, etc)
My 2 cents - JXTA will complement the existing internet and provide an intuitive way to take advantage of P2P computing technology. The best example I can think of is asynchronous web searches that was mentioned in the above URL.
Not sure whether this helps or not, but atleast points to yet another article
Latha
[ February 14, 2002: Message edited by: Latha Kalaga ]

[Kyle Brown]

OK, I dare you to name a protocol in common use that's NOT based on TCP/IP -- does this mean we can expect JXTA over IBM's SNA protocols? WHOO-HOOO! There goes my IBM stock through the roof!
Kyle

[Michael Ernest]

JXTA over AppleTalk? NetBEUI? IPX/SPX? DECNet? Yick.
Now remembering an ugly period at a former job, a PHB proclaiming, "Hooray! They're giving us SNA over TCP/IP!!!"

[Michael Ernest]

Doug -
Out there in the world of network engineers, people want what I informally call AOCOIP (All Our Crap Over IP). These days, Voice-over-IP is finally settling down to a dull roar of available standards (and actual products). iSCSI is picking up steam an an antidote to Fibre Channel, which is nice but expensive and fragile as a physical medium. As part of the technological downturn, it's also proving to be very, very sensitive to market pressure. (That's a nice way of saying the people in the Fibre Channel game are having a tough time right now, and they'd like to customer to keep them flush, even if they can't generate much product at the moment. )
Wiring a conventional Ethernet/IP infrastructure, on the other hand, is commodity stuff. You can buy spools of life-giving IEEE 802.3 wire for cheap, install it without too much careful handling, and off you go. People who do put it in then want to leverage practically everything out of it they can.
TCP/IP, in turn, is well understood, used world-wide (second only, I would guess, to phone traffic), and decent for the bulk of network-dependent applications. You'd want a very, very compelling requirement that forces another approach.
So in the same breath that the spec says JXTA can run over anything you want, it's true and improbable at the same time. What it really says between the lines is: "you can do anything P2Pish you want with this --- as long as you don't mind building it."
Incidentally, Jini is often hyped the same way: "You don't have to use RMI/CORBA -- what freedom!" Well, ok, but what alternatives do I have beyond those two? Whatever I feel like building myself? Wow, now I really do feel free...
[ February 14, 2002: Message edited by: Michael Ernest ]

[Latha Kalaga]

Here is an excerpt taken from JXTA Platform which states

JXTA is designed to be independent of transport protocols. It can be implemented on top of TCP/IP, HTTP, Bluetooth, HomePNA, and many other protocols. This means that a system built on top of JXTA, functions in the same fashion when the system is expanded to a new networking environment or to a new class of devices, as long as there is a correct transport protocol handler for the new networking protocol. The protocols defined in this document can be realized over the Internet, a corporate intranet, a dynamic proximity network, in a home networking environment, or even within a single computer.

Hope that clears up the speculation of AppleTalk / SNA protocols etc.
I welcome any comments from Sing Li on this topic.
Latha

[Sing Li]

Boy, this disucssion is really getting interesting. Thanks for all the comments!
Michael, Latha and Kyle, I appreciate the slight skeptism about �alternate transport support� that accompanies technology such as Jini. For one, I had the very same feeling about the statement in the Jini spec for the longest time � don�t quote me on it, please ;-)
I beg to differ *completely* with JXTA. It is not only probable, but absolutely necessary for the success of JXTA in the medium term!
There is an important reason why multiple alternate network transports support - beyond TCP/IP - REALLY DO make sense with P2P in general, and JXTA in particular. Sorry, gang, but I�m not talking about reviving the popularity of IBM's SNA or Appletalk ;-) It has to do with, surprisingly enough: economics and competition.
There are many examples, but one that comes to mind right the way is wireless devices and wireless networks in general. Most of today�s existing wireless networks use proprietary packet based protocol to transmit data � yes, it is true that TCP/IP can be implemented (sort of) on top of these protocols -- but doing so will require great software contortion and wastage of air-time to accommodate the additional overhead. This means greater COST! Greater cost to fabricate the device for the manufacturer, and greater cost to use for the users.
If you examine the JXTA spec carefully, you will discover that it goes to great pain to ensure the only requirements on the underlying network transport is:
(a)it has a way to address individual nodes
(b)it is able to send a message ONE WAY between one node and another (bi-directional transmission, and reliabile delivery are not even required!)
This literally means that JXTA can be implemented on some of the most minimalist device (no TCP/IP stack and/or emulation required) and over the most primitive of unreliable packet networks. Cost is reduced, and time-to-market is greatly improved.
Unlike Jini, JXTA is designed from day one to support a heterogeneous network consisting of peers using "multiple simutaneous independent network transports" and will (transparent to applications) deliver and route messages through such a network. Finally, networking has a chance of actually becoming pervasive!
It is in this expanded modern day network � one which includes all the �connected devices� in the world, where TCP/IP really shows its age. It will take something like JXTA (if not JXTA itself) to unite all of these devices together on a common computing fabric. This expanded computing fabric is called �Web to the n-th (tm)� by Sun Microsystems and B. Joy � or the �Pervasive Internet� by D. Wang. Thanks, Doug :-)
Best Regards,
Sing Li

[Michael Ernest]

Excellent rebuttal, Sing, and entirely persuasive. I got caught up thinking about JXTA and legacy networks, instead of thinking forward. That's what happens when you get past 35 and don't get the Evangelist job you always wanted.
So, the punishment for being right, yet a larger question: JXTA is sounding more and more open-ended. It's hard to imagine how it will protect itself from turning into another X.500 or CORBA like creature. That is, can we worry that JXTA may become so-all inclusive that the work required to make it happen ultimately inhibits, rather than fosters, choice in the marketplace?
Or is it reasonable to worry that it may become so detail-driven, to ensure compliance among all sorts of platforms, that every product still ends up a little bit noncompliant, and once again we have a products that look portable but really aren't.
Any thoughts on that?

[Doug Wang]

Buddies, thanks for all your responses.
Thanks Latha for your link and quote.
Thanks Michael and Kyle for your arguments clarifying my concepts.
Sing Li, thanks for your informative explanations. But Iam still about the architecture of Jxta network. Could you be kind enough to depict it for us?
I have the same worries with Michael. Yes, �Web to the n-th (tm)� does give us inspiring outlook for Network. But - you said "I do not know of any JXTA compatible mobile agents API/toolkit." And, as Kyle and Michael mentioned, we should have Jxta over AppleTalk, Jxta over IPX/SPX, etc. - all sorts of Transport protocol bindings - at hand. So Jxta has a long run to go (to achieve pervasive networking). Moreover, from Jxta spec, Jxta Project is divided in three layers: Platform, Services and Applications. Its really a BIG cake. Jxta seems to be reinventing everything related to network. Right?
P.S. Jxta seems to be heavily based on XML. What is the role of XML in Jxta?
[ February 14, 2002: Message edited by: Doug Wang ]
[ February 14, 2002: Message edited by: Doug Wang ]

[Sing Li]

Michael and Doug,
To answer both of your questions, and to wrap things up in a reasonable amount of space, I will not attempt to describe the architecture of JXTA (pages and pages) � but give some clues on how things are done and solved by very elegant design.
To alleviate your concern, Michael, JXTA views the various network protocol that it supports as �drivers� (using Doug�s OS analogy). This is very much unlike CORBA, RMI or COM where the connection methods and protocols DRIVES the technology. Because of this, the connection methods and idiosyncrasies of all the different transports are immaterial! Drawing on the OS analogy � the OS views all the different technology CDR/Ws (SCSI, IDE, USB, firewire, 2X, 4X, 16X) in the same way , through the same �driver�.
On top, JXTA provides a very simple and easy to program model consisting of peers, peergroups, and pipes. Any application programmer only need to learn and master this programming model � there is no need to be intimate with the details of every �driver�.
Doug - XML is just "convenient" to use, JXTA does not depend on it in any way -- although if more and more implementations start to use it, backward compatibility may need to be maintained.
<boiler plate follows�>
Thank you, everyone, for a fascinating discussion this week and posting excellent questions. Congrats to the lucky winners.
Anybody who is interested in JXTA should definitely check out the open source site where everything is happening. http://www.jxta.org/ It is where I hang out. There are many commercial companies that are profiled there, complete with links to their websites and description of their products and services.
Of course, my book should also help you to get a quick start on it .
Best Regards,
Sing Li

[Ashik Uzzaman]

I tried to find out why it's named JXTA and got it in FAQ. Sun initiated it and so it starts with J
http://www.jxta.org/project/www/docs/DomainFAQ.html#standfor
Still I am

[Michael Ernest]

Dictionary definition: "the act of placing two things side by side."
Juxtaposition in poetry and rhetoric sometimes describes an unusual combination of two words or phrases.
I take what the designers want to imply is that any two machines (peers) can be placed "side to side" without regard for platform type, OS, etc.