Slashdot Mirror
UCLA, CIsco & More Launch Consortium To Replace TCP/IP
alphadogg writes Big name academic and vendor organizations have unveiled a consortium this week that's pushing Named Data Networking (NDN), an emerging Internet architecture designed to better accommodate data and application access in an increasingly mobile world. The Named Data Networking Consortium members, which include universities such as UCLA and China's Tsinghua University as well as vendors such as Cisco and VeriSign, are meeting this week at a two-day workshop at UCLA to discuss NDN's promise for scientific research. Big data, eHealth and climate research are among the application areas on the table. The NDN effort has been backed in large part by the National Science Foundation, which has put more than $13.5 million into it since 2010.
Just don't expect anyone to early adopt except the usual hypebots and yahoos. We can't even get rid of IPv4 and you want do replace TCP entirely.
This is basically designed to bring the old big media, broadcast ways to the internet. Hence, to basically destroy the Internet, allowing for mass reproduction of centrally created Corporate content, where independant voices are locked out. The protocol is designed for that, mass distribution of corporate created, centrally distributed content to an ignorant, consumption only masses which are treated with disdain and objects of manipulation by the elite. This is to bring big media and the stranglehold they had for so many years on information the public has access to back.
With the Ipv6 transition needed its time to focus on that rather than on this plan to destroy the internet and turn it into the digital equivalent of 100 channels of centrally produced, elite controlled, one way cable television programming designed to psychologically manipulate and control a feeble and dim witted public.
No thanks and get your #%#% hands of my internet.
They are also funding a study to replace roads with run-flat tires. Oh, right, different layers.
I was puzzled with the involvement of Tsinghua University of China with this thing
After reading your comment it starts to make sense
The China Communist Party needs to regain control of the Internet (at least inside China), that explains why they endorse this new scheme so much
Muchas Gracias, Señor Edward Snowden !
Unfortunately, as we learned from the debacle of cellular communications, corporate inertia will either squash this or slow gestation until it's stillborn. There is a substantial investment in the current technology of TCP/IP and it still works "just good enough". This change in network would require installation of a twin network alongside the current, with slow adoption in the consumer side. That would be very expensive to build and maintain over numerous financial quarters and thus no MBA-centric company would ever do it in current corporate culture. This takes long-term thinking in a quarter-to-quarter environment. Thus it won't happen for a very long time.
I have something in common with Stephen Hawking...
TCP/IP has the singular advantage that it is deeply entrenched, runs on a vast number of devices from supercomputers right down to single-chip computers. Is it perfect? Absolutely not, but it's a proven technology.
I'm sure in the fullness of time it will be replaced, or at least subsumed into some better protocol, and maybe this initiative will be the one that produces its successor... or not. I think TCP/IP is going to be with us for a very long time.
The world's burning. Moped Jesus spotted on I50. Details at 11.
Despite decades of research the horse and cart are still the best thing we know for the task at hand. Yes, it's admittedly not really good, but all the known alternatives are worse. This is more likely some kind of publicity stunt or serves some entirely different purpose.
Your statement as shown can be applied to the internal combustion engine, or any other technology. Rejecting any change out of hand without consideration is incredibly sad, if not dangerous to our species future prospects. Yes it's important to take everything with a grain of salt, but everything should be at least considered. It only takes one successful change to have a dramatic impact and improve the lives of many.
This goes for all technology, not just this specific problem.
There is a talk on youtube from 2006 by Van Jacobson that describes this idea before it was called named data networking. It is really neat, and I am surprised that it has taken so long for somebody to actually try to implement it.
http://www.youtube.com/watch?v=oCZMoY3q2uM
Don't worry the NSA and GCHQ interests are being covered by China.
i thought once I was found, but it was only a dream.
From the architecture page:
Note that neither Interest nor Data packets carry any host or interface addresses (such as IP addresses); Interest packets are routed towards data producers based on the names carried in the Interest packets, and Data packets are returned based on the state information set up by the Interests at each router hop
Great, NAT-like state in every router...
BTW, how hard will it be to transform Linux's kernel structure into something that is equivalent to Plan-9?
not very.
http://www.glendix.org/
http://en.wikipedia.org/wiki/P...
http://en.wikipedia.org/wiki/I...
---Saying gnome 3 is better than windows 8 not so much a compliment as it is damning with light praise.
First, IPv6. If you can handle simple things like that, then we'll let you play with the important stuff.
Oh yeah. Flying cars too.
Have gnu, will travel.
All the internet is NOT "give me data named thus." For example, this "NDN" doesn't seem to support logging in to a particular computer, you know, so that you can administer it. It doesn't seem to support sending a file to a particular printer. Maybe it might make an interesting overlay on IP, replacing existing content distribution techniques, like Akamai, but I'm not seeing it replace IP.
-- david newall
I'm glad they are starting this now so hopefully by the time we run out of IPv6 addresses, we'll be ready!
We do not know whether there is a better solution, but currently we do not have one, despite decades of research. What would you do, start breaking things?
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
How is this going to harm the everyday Internet user? I imagine at the very least it will make it more difficult for two random internet users to connect to each other, because all connections will probably have to be approved by Verisign or some other shit like that.
Remember folks, the age of innovation is over. We are now in the age of control and oppression. Everything "new" is invented for one purpose and only one purpose - to control you more effectively.
I could totally see the two networks running simultaneously. It's completely accurate that TCP/IP sucks for mass content delivery; it's gigantic waste of bandwidth. And for point-to-point interaction this protocol would be massively inefficient.
But why can the two protocols not run on top of the same Layer 2 infrastructure?
In a nutshell, this is applying DRM to all of your connection attempts. You will only be able to make connections that are "authorized" by TPTB.
No more free and open networking.
Since every single goddamned one of you has used magnet links, you should be comfortable with the idea of requesting objects rather than discussions with particular hosts. Taking this idea and running with it is NDN. It's an excellent network research subject.
It facilitates caching, multipathing... with some more work perhaps network coding to get close to the min-cut bound. Bittorrent is super successful because it's all about the content. Let's give a similar protocol a chance at changing the net.
Comment removed based on user account deletion
After reading the spec, it seems to me that this is a collapse of the HTTP (web) protocol down to the network/transport level. In effect, the internet would become one large heirarchical namespace where clients ("consumers") query the heirarchy of data by uri through Interest Packets and then some server somewhere sends back a Data Packet matching the specified interest. Alot like 20th Century TV, sounds like.
Also there is a provision for packet signature using public-key RSA which makes me think that it would be easy to instruct internet routers to deny passage for all packets not coming from or going to officially sanctioned sources/destinations should the need arise.... makes my paranoid little brain somewhat nervous....
This seems like a fun project to implement over TCP/IP, rather than to replace it. Afterall, there's no shortage of fibre (bandwidth) we could lay, so it makes little sense to abandon the "any peer to any peer" model of the current internet for one that might be better organized just to conserve bandwidth.
The more intelligence we put into the routers and network/transport protocols, the more the internet could start to resemble the old-style telephone company (or cable TV), where the devices on the edges of the network (meaning us) have very little.... creative legroom. This is something that I think we want to avoid.
For those who don't see why this is bad, consider this:
In order to route/cache by data, the data must be visible to the routing nodes; in essence, you would no longer be able to use end-to-end encryption. You could still have point-to-point (eg: encryption for wireless connections), but everything would be visible to routing nodes, by necessity. This means no more hiding communications from the government (who taps all the backbone routers), no TOR routing, no protection from MTM attacks, by design. You get the promise of more efficiency, at the cost of your privacy/freedom... and guess what, you'll get neither in this case, too.
Why should content protection be part of the Internet standard? Why do my devices (routers, computers, etc.) have to have built in DRM which will end up getting cracked, or at least possibly exploited from offshore?
This also is going to be met with a lot of suspicion. Who keeps the keys, gets to keep content locked, owns the license servers, and is able to come in via backdoors mandated as part of the protocol? The UN? Give me a break. China? Sure, we can trust them allright, provided we give them 51% ownership of any venture. It won't be the US because BRIC will sooner create their own network and completely split off.
I don't reject change... but what does this new protocol give me? IPv4 and to a lesser extent IPv6 have been torture tested, are completely open, and one can cobble together adequate defenses against attacks not too expensively (Cisco ASAs on the low end are a couple C-notes, and there are always smaller routers). A protocol based around DRM and content protection, stuff that is made to obfuscate and lock down is not going to be of any benefit to anyone but a few.
To boot, this seems like a complex mess. A network protocol should be brain-dead simple in order to reduce the attack surface, and reduce bugs. Adding DRM at layer 2 is at best will slow things down, at worst, allow the bad guys to hide behind bogus certificates.
Grabbing my tinfoil hat, I'm wondering if this protocol is something that will end up mandated within hours as soon as a "warhol event", or something more known as a "cyber 9/11" happens. I would not be surprised if this is already written and ready to be thrown on the floor as a bill on both houses the second some major security breach happens that causes catastrophic damage.
I'm seeing shades of the Clipper chip again, with the same problems. The bad guys getting access to the backdoors, compromising everyone in a way that cannot be patched, the bad guys closing the backdoors so they can't be investigated by LEOs... and the biggest losers are the good guys.
I've used SCTP. It's not particularly better than TCP. It has some things that make it nicer if you are doing all your programming by writing directly to the socket.
But no one actually does that. In practice, even people writing low-level code encapsulate their send/receive in a function or a method, at which point SCTP doesn't give any real advantages. The idea of channels is kind of cool, but for it to be really useful, they would need guaranteed bandwidth (or once again, encapsulating your network code in functions will give you the same result with TCP).
Add to that, the kernel driver code for SCTP isn't well tested (because it's not well used), and SCTP is really a solution looking for a problem.
"First they came for the slanderers and i said nothing."
The advantage of SCTP is that it is not a retarded implementation of go back N.
SCTP has all the same limitations as TCP at the SCTP stream level.
Which means it can operate efficiently at high speeds on unreliable networks. Also the channels could be easily and automatically used with HTTP to replace the inefficient pipelining. With TCP something like SPDY had to reimplement channels on a higher level.
This is semantically identical to opening multiple TCP sessions - one for each stream. If you were to lower round trip cost of subsequent session setup in TCP to zero (e.g. fast open extensions) then you essentially have the useful advantage of SCTP without SCTP.
The only benefit SCTP has is multipath failover baked in and you can't even use the extra paths concurrently it only exists as a contingency.
It looks like they started out with Content Addressible Networking, which is a great idea. Massive bandwidth savings, improved resilience, faster performance, power savings, everything you could want. But then rather than try to impliment CAN properly alongside conventional networking they went for some ridiculous micro-caching thing, over-complicated intermediate nodes that enforce usage rules, some form of insane public-key versioning system validated by intermediate nodes and generally ended up with a monstrosity.
CAN is a great idea. NDN is a terrible implimentation of CAN. The main selling points include having DRM capability built into the network itsself, so if you try to download something not authorised for your country the ISP router can detect and block it. A simple distributed cache would achieve the same benefits with a much simpler design.
There's the core of a great idea in there, burried deep in the heap of over-engineered complexity that appears designed not to bring benefits to performance but rather to allow ISPs to readily decide exactly what content they wish to allow to be distributed and by whome. This thing is designed to allow the network devices to transcode video in real time to a lower bitrate - putting that kind of intelligence in the network is insane!
NDN looks like a scheme to tag data and change networks from "addressing a particular node" to "addressing data". This is like changing the Post Office such that a person addresses a particular letter sent to them, rather than having a house number where letters get delivered.
Computer addresses with DNS on top make sense: it's easy to subdivide and route, and name translation allows humans to interact with it. NDN looks like it's trying to make the names the addresses, and make the URIs the names, and make the routers act as caches, and hope it all works; but then how do I address a *computer*? How do I ask for anything other than HTTP?
NDN looks like p2pwww stuff I designed back in 2004, except trying to implement as a network protocol on the routers, rather than an application protocol on the nodes. Even then, I specified digital signatures, encryption, and network namespace isolation: you could have an ICANNWeb which signed certificates for each name (i.e. Microsoft) and, on ICANNWeb, you would put out a message (P2P) for Microsoft://www/windowsxp/support.aspx and get back responses for (have|know|home)--node has a copy recent as per [date], node knows who has a copy recent as per [date], node knows the home is [address]--and select from there. Each resource would be digitally signed with generation date stamp and expiration date stamp, and a new generation date stamp overrides an earlier expiration date stamp.
In short: you'd get on a Gnutella-like network, perform a search, and be told where the resource is. Data was such that you could identify newer, identical, and expired resources. Your node could say, "0-3 hops", then "4-6 hops", incrementally crawling the network; or "3 hops past first response, limit 10". Usually if a node knows another node has a copy, that other node also knows several (it got its copy somehow--by its own request). If a node locates nodes with multiple versions, it provides outdated nodes with provable evidence that they're outdated, so they can drop their caches and learn some other node has a more up-to-date copy. Likewise, when those nodes are queried, they will then re-query the nodes they know have copies, and update them: an update doesn't trigger this cycle--too much traffic.
That's application-level. A locatable, self-caching network which encapsulates all resources in digital signatures and allows for namespaces. It sounds like that's what they're trying to accomplish, but in the transport layer.
Support my political activism on Patreon.