Researchers Scheming to Rebuild Internet From Scratch

BobB writes "Stanford University researchers have launched an initiative called the Clean Slate Design for the Internet. The project aims to make the network more secure, have higher throughput, and support better applications, all by essentially rebuilding the Internet from scratch. From the article: 'Among McKeown's cohorts on the effort is electrical engineering Professor Bernd Girod, a pioneer of Internet multimedia delivery. Vendors such as Cisco, Deutsche Telekom and NEC are also involved. The researchers already have projects underway to support their effort: Flow-level models for the future Internet; clean slate approach to wireless spectrum usage; fast dynamic optical light paths for the Internet core; and a clean slate approach to enterprise network security (Ethane).'"

Comments

The Six Million Dollar 'Net.

[mikecardii]

Gentlemen, we can rebuild it. We have the techonology. We can make it better, faster, stronger.

Re:The Six Million Dollar 'Net.

[Fyre2012]

We can make it better, faster, stronger.
Now if we could only rewrite Windows from the ground up :)

Re:The Six Million Dollar 'Net.

[kaizenfury7]

....and with DRM baked in.

Re:The Six Million Dollar 'Net.

[KingSkippus]

Now if we could only rewrite Windows from the ground up

Didn't you see the story the other day?

We are.

Re:The Six Million Dollar 'Net.

[pipatron]

Funny, that was exactly what I thought even before I read the summary. I bet there will be no chance to browse anonymously this time.

Re:The Six Million Dollar 'Net.

[cayenne8]

"Gentlemen, we can rebuild it. We have the techonology. We can make it better, faster, stronger."

Unfortunatly, I'm afraid they will make it more censorable, more business oriented vs. regular people, less anonymous, more regulated, govt/UN controlled, politically correct...and as someone mentioned, full DRM support forever.

Frankly, for all its faults, I like the internet now as it is...kind of the 'wild west' of information. That just has to 'kill' some of those in power around the world.

I think the last thing we want to do, is recreate it, now that those in power know what free flow of information can do...

Re:The Six Million Dollar 'Net.

[mrchaotica]

I think the last thing we want to do, is recreate it, now that those in power know what free flow of information can do...

Indeed, the only way to "recreate" it is to make it even more decentralized and unregulated!

Re:The Six Million Dollar 'Net.

[westlake]

I like the internet now as it is...kind of the 'wild west' of information.

The "Wild West" exists (and perhaps always has existed) mostly in fiction.

In history it begins with the discovery of gold in California in 1848 and ends in 1876 at the Little Big Horn. The Last Stand for the Plains Indians as well as for Custer.

It's a brief moment in time - and, in some ways, a pattern of settlement unique to the United States.

It shouldn't surprise anyone if the Internet frontier has it's own ending.

Re:The Six Million Dollar 'Net.

[GringoCroco]

From the whitepaper PDF:

It should be:
1. Robust and available
2. Inherently secure.
3. Support mobile end-hosts
4. Economically viable and profitable.
5. Evolvable.
6. Predictable
7. Support anonymity where prudent, and accountability where necessary.

Re:The Six Million Dollar 'Net.

[sehlat]

I think the last thing we want to do, is recreate it, now that those in power know what free flow of information can do... Damn straight. If the Powers That Be had seen what an open network would do, they'd have strangled it in its cradle. Quite possibly, in their "generosity," we might have gotten a centralized "information utility" monster, something like France's "Minitel" system on steroids, with all information filtered, censored, corporatized, and source-trackable. Feh.

Re:The Six Million Dollar 'Net.

[trianglman]

7. Support anonymity where prudent, and accountability where necessary.

Who determines necessity? If left up to any current government, the necessity would be determined by who wants to be anonymous. Senators - sure, they need privacy for their solicitations of pages; Joe Shmoe Public - nah, its better to keep tabs on him, he could be a terrorist...

Re:The Six Million Dollar 'Net.

[lanc]

A rewrite/new tech doesn't always mean real-life solution. See OGG vs. MP3.

Re:The Six Million Dollar 'Net.

[cheater512]

IMHO we should give it a chance. The internet has plenty of problems including scaling, spam, etc...
If they could be fixed then it would be really good.

If they start limiting free speech and the like then we can stop supporting them.
Whats the point of having a new internet if noone is going to use it?

Re:The Six Million Dollar 'Net.

[beckerist]

From the article:

The program will collaborate with, and be funded by, approximately seven industrial partners with interests in networking services, equipment, semiconductors and applications.

...I would assume that it would be determined by those who are building it....meaning ^^ those guys!

Re:The Six Million Dollar 'Net.

[itlurksbeneath]

Umm.. Judging by how much infrastructure would have to be replaced, I think you were off by a factor of a million (and I don't mean the 6 dollar web, either).

Re:The Six Million Dollar 'Net.

[Archangel Michael]

I actually fear my Senator more than Joe Shmoe Public. Congress is filled with TERRORISTS in my book.

1) Vote for me, I won't take away ________ (government subsidy) _______ (Political Opponent) willl!

2) Vote for me, I'll protect you from _______(enemy of the state), _________(Political Opponent) won't!

It is all "Terror" based politics.

Re:The Six Million Dollar 'Net.

[kad77]

What exactly is terrifying about losing a particular government subsidy?

Get off the tit! Many "artificial" subsidies need sunsetting at some point in time.

A person/party that has a proven track record of not being practical about core functions of government like the safety/security of its people would be a far more worrisome prospect than who gets the cheese...

Some day... well, maybe not..

Re:The Six Million Dollar 'Net.

[willy_me]

I would support a solution where limited anonymity is supported for everything. What I mean by limited anonymity is that anonymity is provided but the ability to bypass it is provided if required. Doing so would be like tapping phones - a warrant should be required. The exact rules would change depending on what country you are from.

Who determines necessity?

In this case, the lawmakers and justice system. In a democratic nation, the people would have the final say (in theory).

Joe Shmoe Public - nah, its better to keep tabs on him, he could be a terrorist...

A good point, but not really one that concerns me. I'm no terrorist and I don't care if the government watches over me. But if that information was used for other purposes, then I would get upset. I believe requiring a bit of paperwork and convincing a judge to grant a warrant would provide an adequate means to limit the temptation for abuse. It's really the corporations that I want to keep my information private from and they aren't in the position to gain access to such information.

It's all about compromise, just like everything else we have in our society. We are not truly free, we live by rules so that we can all live together. The same applies to any sort of shared environment - including the internet. Adopting the correct rules for a new internet will allow everyone to benefit. So rules (or the ability to implement those rules) should not be discounted outright, it should be assessed and only implemented if beneficial to the community as a whole.

Willy

Re:The Six Million Dollar 'Net.

[drakaan]

I liked the whole "economical and profitable" requirement...nothing like mentioning QOS/bandwidth throttling and economic viability in the same paper to get the backbone providers to pony up some research money...so long net neutrality, we'll miss you.

Re:The Six Million Dollar 'Net.

[alphamugwump]

You kind of picked a bad example, as vorbis is actually quite popular in some domains (games, for example), is supported by several hardware players, and gives better compression than mp3.

Of course, the idea of rebuilding the internet is a load of bull. The article lists a bunch of things you supposedly can't do with regular protocols, and takes those as reasons for change. They seem to think we can't do multicast, QOS, or security with current protocols. They also seem to think that, since wireless is so different from land lines, we should need new protocols. Their plans also happen to destroy any possibility of network neutrality.

I sincerely hope this project doesn't get any government funding.

Re:The Six Million Dollar 'Net.

[MetalPhalanx]

I think that he wasn't so much making a point about losing a particular government subsidy, but that instead of improving themselves to look better than the competition, they are just trying to make the competition look worse.

Re:The Six Million Dollar 'Net.

[EaglemanBSA]

...but I don't want to spend a lot of money.

Re:The Six Million Dollar 'Net.

[Anonymous Coward]

Frankly, I can't even begin to take notice of anyone with such a crap website!

Re:The Six Million Dollar 'Net.

[mr_mischief]

It's a reference to an old TV show, "The Six Million Dollar Man". Lee Majors played Steve Austin, who got terribly injured. The Powers That Be rebuilt him as a bionic wonder able to do things no mere flesh-and-blood human can do. The doll^H^H^H^Htelve-inch action figure had a vamera view-finder type of thing that you looked through his head to use (his bionic vision), and a cool karate chop action arm.

Re:The Six Million Dollar 'Net.

[itlurksbeneath]

I'm aware of that. I watched the show when it came out.

Re:The Six Million Dollar 'Net.

[jhaar]

FYI: it's "better, stronger, faster"

(dee-doop, dee-dee, dee-doop, dee-dee, DA DA DA DA DAAAAA!....)

Re:The Six Million Dollar 'Net.

[Archangel Michael]

Tell Granny she is going to lose SS and see what happens to Granny's vote. She is going to vote for the guy that say he'll protect SS. It is nothing more than Terrorism, but instead of guns to women and children, and blowing up delis, it is financial.

Same with the other side of the court. Tell Granny that the other guy wants illegal mexicans coming in to rob and rape her, and see who she is going to vote for.

Re:The Six Million Dollar 'Net.

[Anonymous Coward]

I wouldn't say it ended in 1876. Things got pretty wild in Tombstone, Arizona in 1880. The Gunfight behind the OK Corral in the empty lot by Fly's Photography studio was just one incident in a series of "wild west" activities that occurred then.

Re:The Six Million Dollar 'Net.

[AGMW]

I'm no terrorist and I don't care if the government watches over me.

If you haven't done anything wrong, you've nothing to fear, right?

Re:The Six Million Dollar 'Net.

[GuyWithLag]

This is the typical academic boil-the-oceans scenario, and I doubt it would ever work - we can't even migrate to IPv6, which is backwards-compatible with the current setup, while theirs isn't (by design, no less).

Re:The Six Million Dollar 'Net.

[dpilot]

Anonymity is really, really thorny, and to pretend it's anything else is to deny reality.

The good side of anonymity:
The ability to seek information without fearing repercussions for having sought it.
The ability to give "anonymous tips," again without fear of repercussions.

The bad side of anonymity:
DDOS and other general nastiness.
The ability to give intentionally incorrect malicious "anonymous tips" without fear of repercussions.

There's far more than just those, but even those are powerful arguments on both sides of the issue.

As for the "where necessary," I was at a security-oriented meeting last week, and brought up the fact that we don't really have a specific legal framework in place for key escrow. What's to stop Joe Shmoe from setting up "Joe's Key Escrow Shop" and what sets up his relationship to the Law, police requests, subpoenas, etc? Then again, if key escrow were to be split into 2 agents, where's the legal framework to make sure those 2 agents are kept legally and financially independent, etc? The issues of an "identity broker" necessary for "anonymity where prudent, and accountability where necessary" are pretty much the same as key escrow.

Re:The Six Million Dollar 'Net.

[rasputin465]

If they start limiting free speech and the like then we can stop supporting them.

The whole point is that this feature [limiting free speech] is in their design plan, so you can `stop supporting them' now.

Re:The Six Million Dollar 'Net.

[trianglman]

I'm no terrorist and I don't care if the government watches over me.

Unfortunately, that doesn't help those in governments where saying the wrong thing to the wrong person can get you locked up without a trial. Similar things have happened to a couple American citizens (and people unfortunate enough to have been noticed by the American administration, accidentally or otherwise) in America. Your innocence is only a protection if those who would persecute you need to prove your guilt. It does nothing when you are never even given the chance to prove your innocence.

It's all about compromise, just like everything else we have in our society. We are not truly free, we live by rules so that we can all live together.

Agreed, but in order for there to be compromise, those who have to follow the rules should be given equal voice to those who want to set the rules. When laws are being made by people who know little about the subject matter (a series of tubes anyone?) and those people are elected by an even less informed populace, you aren't going to get a compromise that is going to help you or me. You are more likely to get one that helps AT&T and the NSA to continue to monitor everything you do online.

Re:The Six Million Dollar 'Net.

[Anonymous Coward]

In history it begins with the discovery of gold in California in 1848 and ends in 1876 at the Little Big Horn.

You must be from "back East." Out here in the west, we have a different take on that.
The Lincoln County war in New Mexico started in 1877 and could be said to have ended with the death of Billy the Kid in 1881. But that wasn't the end of the wild west, either.

While the great transcontinental railroad was finally connected in 1869, major railroad construction continued into the 1930s, and railroad camps in the west were much in the tradition of the wild west.

I think you could say that the wild west mentality continued through the Depression and the make work camps of the 1930s. It might be said that the Wild West ended with WWII.

But then, the street violence of today still reminds us of the Lincoln County war of the 1870s.

Re:The Six Million Dollar 'Net.

[Anonymous Coward]

Assuming they are going to go with something like 7 layer OSI model, they are going to have the Physical layer responsible for "Support anonymity where prudent, and accountability where necessary". Wow! So I can't get my new Network card from just about anywhere (such as Fry's) but only from Stanford University bookstore.

Re:The Six Million Dollar 'Net.

[beckerist]

I think it's not out of the realm of reality to have a project financed privately, yet retain objectivity in the project itself. Just because it's optimized for banking, QoS, VPN's, encryption and accountability DOESN'T mean that there couldn't be some sort of "first amendment" law, or some sort of encrypted anonymous method to browsing... I agree that it wouldn't be truly "open" anymore, but I don't think that's any worse than say, the move from POP mail to webmail. Sure you can use the same programs (so long as they support the interoperability,) and yes, there were things you could do with POP mail that are difficult to do with webmail (simply because those programmers who made the webmail application might have not chosen to ALLOW all of the functionality one would see in a POP client.) One could argue this is due to security, privacy, power...but one could also argue that it's so that Google can scan your mail and advertise to you appropriately. Bad? No. Annoying? Maybe, but I'm willing to live with it if it means I get to use a tool like Gmail (especially for free!)

Re:The Six Million Dollar 'Net.

[drakaan]

I understand, but I have huge problems with most of the idea. Here's why:

First, we must harken back to the issue of public money going to the telcos to build out high-speed networks. More than a handful of billions of dollars spent there by everyone went straight into the coffers of the telcos, and we haven't seen the fruit from whatever effort may have been made there.

That ties in with the whole tiered-service/degraded service offerings that the big backbone providers seem to think are such a good idea. Yes, the fact that there's not currently a statute prohibiting service degradation hasn't meant a complete slide into pay-through-the-nose type service, but that's primarily because under the current infrastructure, it's more difficult to implement large-scale and fine-grained QOS controls.

What we're talking about now is an institution trying to get funding (private funding) for developing a network that would supplant the existing internet infrastructure (and presumably make use of the same fiber...economical and all) and allow for robust, built-in management of bandwidth, service quality, etc.

Adopting the "internet as tubes" mindset (which this goes a long way towards making a reality), those who own the "pipes" can more easily open or close particular valves.

I rather prefer the "internet as a superhighway" metaphor, even though it may come with its share of traffic jams. Military and emergency service concerns don't especially change my mind...in both instances, a completely separate network seems a better solution. We've all paid for the roads...repeatedly...let's remain agnostic as to just how much traffic of a particular type is allowed on them (or how much a particular user is allowed to use them).

Going from POP/SMTP to HTTP as a user interface for email doesn't change the underlying way that mail gets transmitted from one mail system to another. It was developed because browsers got smart enough to make themselves become reasonable and cross-platform mail clients. I'm all for web interfaces...it's the lowest common denominator for application interfaces these days, but not all new and exciting internet software is sitting in a web page.

The freedom to innovate (especially for those of us who do network programming) that the current internet provides shouldn't be underestimated. There's a lot to be said about being able to expect equal treatment for your brand-new, whiz-bang protocol or application that you just don't see until somebody decides that it should be treated as a second-class citizen (e.g. VOIP).

If I want to be able to let people connect to their inboxes via SMTP, POP, HTTP, or IMAP, I should be able to (provided I paid for an internet connection). If I want to provide the same via RSS, bittorrent, or gnutellanet, I don't want to be at the mercy of those that I have paid for internet service from. I'm not paying for HTTP, FTP, and GOPHER connectivity, I'm paying for an internet connection.

We've already been moving further and further away from the internet as a peer-to-peer network and towards the internet as Compuserve/Prodigy/AOL for some time. In the interests of making it a worthwhile thing to have in existence at all, I'd just as soon see the internet remain as neutral as possible. If this continues, either BBSes, FIDONet, and ToadNet might be seeing a big resurgence in popularity, or everyone will be trying to figure out how to run new services on top of whichever ones are least likely to be throttled (SMTP over HTTP, Bittorrent over RSS). The network is *supposed* to be dumb, just like a highway.

Re:The Six Million Dollar 'Net.

[cburley]

The good side of anonymity:[...]

Great points, from the political/social/morality view.

From a technical point of view, to assess anonymity, as a feature and/or a bug, one must look closely at the underlying transport protocols and mechanism.

If that mechanism does not inherently (or intrinsically) identify an entity, then allowing anonymity in higher-level protocols is a feature; otherwise, it may be a bug, in the sense that insisting on providing it may well introduce bugs.

Expanding on the former, it's important for any communications protocol built on top of TCP/IP to not require identification of an entity above and beyond what TCP/IP inherently provides (mainly, IP addresses of the communicating entities).

Otherwise, all sorts of problems arise involving those higher-level identities, such as forgery and additional points of failure. (The only way to be sure a given entity is who it says it is is to check an external data base, which might not be available or might be buggy.)

Expanding on the latter, it's difficult to provide anonymity in a protocol when the lower-level protocol requires identification to work. E.g. it's difficult to anonymize one's IP address when communicating via a protocol built on top of TCP/IP.

Again, such anonymization can be done, but difficulties are encountered beyond the non-technical ones, such as additional points of failure (an anonymizing agent might be unavailable to facilitate communications between two entities that are both available and able to communicate directly with each other) and bugs introduced by unexpected reliance on correct identification in underlying and/or related protocols (which is why, for example, the FTP protocol is often "special-cased" in stateful firewalls, NAT boxes, etc. -- it doesn't rely solely on a single established TCP connection to work, it also uses IP addresses, or identifiers, in ways that have to be tracked and potentially modified by anonymizers/translators).

So, the "new Internet" I want to design is one that relies less, not more, on centralized agents (such as data bases, including DNS) and brittle assumptions (such as reliable voluntary identification of entities), in comparison to today's Internet -- not so much for political or social reasons, but because I believe the result would be a more viable, robust Internet.

On top of that, I believe it would actually be easier to add support for the "usual suspects" (encryption, DRM, whatever) as desired by entities that actually are willing to employ them, in line with the end-to-end principle, leaving "most" of the rest of the Internet out of those decisions and, therefore, letting them avoid the potential problems that arise from having to support them for everybody, all the time, even when they're not desired for certain communications.

Damnit

[0racle]

I haven't even upgraded to Internet2 and Web 2.0 and they're already doing work on Internet3.

Re:Damnit

[User 956]

I haven't even upgraded to Internet2 and Web 2.0 and they're already doing work on Internet3.

They're not just working on Internet3. They're working on Internet360, which is 120 times better than plain old Internet3.

Re:Damnit

[saskboy]

Maybe we could have email then, that doesn't tell me that my penis is too small, and needs C1alis. Because, like, those are both totally not true. I blame the Internet.

Re:Damnit

[Anonymous Coward]

I haven't even upgraded to Internet2 and Web 2.0 and they're already doing work on Internet3.

Me neither. I'm waiting until at least Internet XP and may even hold out until Internet Vista with the built-in DRM (Digital Rights Manacles). Of course, then I'll have to buy all new hardware in order to run the Internet at full resolution.

Re:Damnit

[Slipgrid]

They're working on Internet360, which is 120 times better than plain old Internet3.

Or, 120 degrees in the wrong direction.

Re:Damnit

[ImaLamer]

Think of this more as Internet 1.5

Re:Damnit

[saskboy]

And here I thought we were already running Internet 7

~ Some Microsoft noob

Re:Damnit

[webheaded]

No you see, they're going to make a complete 360! They're going to add all kinds of stuff to the internet and when they're done, they'll be right back where they started, just like the XBOX 360!! Yeah!!!112

Re:Damnit

[syrion]

Only in Serial Experiments: Lain.

Hmm..

[chowder]

Is someone going to call Al Gore and get his opinion on this?

Re:Hmm..

[whorapedia.com]

Oh right - because he invented the internet... hahaha... good one, jackass

Re:Hmm..

[choongiri]

Pssst... I think you may find this page informative:

http://en.wikipedia.org/wiki/Joke

Re:Hmm..

[Anonymous Coward]

Only a dumbass would think that that joke was funny, jackass.

Re:Hmm..

[Anonymous Coward]

I think I love you

Re:Hmm..

[syphax]

I daresay that if we collectively hadn't had so much fun mocking Al Gore up to November 2000, we just might be in a better place today.

I'm not a big Al Gore fan, but JFC we are a lot closer to 1984 under the current administration than I ever would have imagined. It would be hard to do much worse.

Re:Hmm..

[fredrated]

If you don't think George Bush is a bigger joke (read catastrophe) than Al Gore then God help us, no one else can at this point.

Re:Hmm..

[Zantetsuken]

yes, the people who upgraded his invention of the net to use *BOTH* trucks *AND* tubes - all so that they can file a new patent and bring on the lawsuits down on Gore for the patent they filed 5 minutes from now...

Re:Hmm..

[syphax]

Flamebait? Maybe.

Offtopic? I beg to differ. Grandparent illustrates the trap that many of us fell into in 1999-2000: It's fun to make fun of Al Gore!

It was fun, it was easy, and it had consequences. You may remember that the 2000 election was rather close.

Re:Hmm..

[Anonymous Coward]

l2humour

Sounds great...

[cedricfox]

...but the biggest hurdle is convincing people not to connect to these shiny new networks until it's all in place, end-to-end. It seems like this would have to be physically secured while it is being put together.

Re:Sounds great...

[Tackhead]

> ...but the biggest hurdle is convincing people not to connect to these shiny new networks until it's all in place, end-to-end. It seems like this would have to be physically secured while it is being put together.

Oh, that's simple. Don't put any pr0n, MP3z, movies, or warez on it until it goes live. Then, unleash the .torrents of hell.

Re:Sounds great...

[fmobus]

Why? I'd rather let it grow "organically" than getting everyone online at once. Nevertheless, I think the burden would be interfacing legacy stuff (the tubes we are using now) with the shiny new stuff consistently.

Re:Sounds great...

[Anonymous Coward]

Exactly what I thought when I saw this news. The good thing about the internet is its contents. To get people to use the new internet, you would have to bridge it with the old one and let people access old and new internets at the same time. But that would compromise the security, wouldn't it?

Re:Sounds great...

[cedricfox]

Well yes, but then you have the new fast Internet with its own security protocols, and they're compromised, de facto, by being attached to the old Internet and its legacy protocols and modes.

What material will they use?

[Recovering Hater]

Are they going to go with rigid copper, flexible copper or PVC? It is just a bunch of tubes right?

Re:What material will they use?

[Tokimasa]

Maybe pneumatic tubes. Or carbon nanotubes. Or inner tubes.

Re:What material will they use?

[xerxesVII]

Inner tubes. It is the Innernet, after all.

Re:What material will they use?

[EinZweiDrei]

I'm thinking manicotti.

Re:What material will they use?

[voice_of_all_reason]

Those transport ones in Futurama looked pretty fun.

Re:What material will they use?

[dubbreak]

It is just a bunch of tubes right?

Actually they discovered the problem is that the current internet is a bunch of tubes. Tubes get clogged. The new internet will be big trucks you dump stuff on.

What are the odds

[Lokatana]

What are the odds that, even given a great plan, that this has any hope of making it to daylight. IPv6 has been out for how long, yet how much real adoption have we seen in that space?

Re:What are the odds

[Anonymous Coward]

IPv6 is an incremental change to one component of the entire communications structure.

CleanSlate is apparently about stepping back from the entire structure and asking how we could best reach end goals of communication (e.g. sending a secure text message to a colleague, not "how to encrypt Email better").

And this is done with lessons learned in mind along with all the mathematical techniques we've honed since internetworking began.

Do the project initiators think their ideas will be put into practice? Who knows. I bet they don't either...yet. Looks like a grand thought experiment which may inform future, specific design in small areas or possibly-but-probably-not result in a well considered, complete replacement design for a new global networking structure.

I'm just happy some bright minds are going to put some serious time thinking about and discussing the idea.

Re:What are the odds

[SEAL]

The only way such a project will succeed is if it can be implemented side-by-side with the existing infrastructure, and then provide a migration path to keep costs down for companies. If the hardware changes are minimal, and their existing business is not impacted, then companies may make the jump. However, part of it is a chicken-or-egg problem. Companies don't want to migrate to an arena with no consumers, yet consumers don't want to use an Internet with no stores.

There would have to be additional, compelling benefits to switch. Say, for example, the existing Internet continues to degrade in performance under the burden of spam and DDoS attacks. If the new network provides built-in avoidance of those issues, then the incentive to use it is stronger.

Re:What are the odds

[griebels2]

The problem of IPv6 is due to the fact that it just doesn't work besides IPv4. You essentially need to build and maintain two seperate networks. Yes, you can share the same equipment, but the amount of configuration involved almost never justifies the efforts in corporate environments.

In my opinion, there are a lot of things that need to be fixed for an "Internet for the future". One of the biggest hurdles of course is the address space shortage of IPv4, but there are a lot of other issues which need to be solved. Just to name a few:
- More flexible routing of unique identifiers (let's call them IP numbers), so I can take my "identifier" with me (think mobile phones)
- A solution to the ever growing "global routing table" (BGP4 as it is used today)
- Better support for quality of service from end-to-end.
- Better "multicasting" support, also end-to-end. (Let's avoid burning down networks during "cataclysmic" events)
- Better redundancy. Although dynamic routing protocols should heal this problems, in practice they often fail to do this. Especially in cases where connections are semi-dead)
- A much better built-in protection against DDoSes and other kind of abuses.

Unfortunately, IPv6 really fixes none of those problems, except the IP number shortage. IPv6 also comes at great costs, since you need to upgrade your whole infrastructure at once, or it isn't really usable.

So, IPv6 might have been a nice lesson for the next generation "IP protocol". IMHO this next generation should take the following things in mind:

- Transition only works if it plays nicely with the legacy stuff during the transition.
- Transition has either to be cheap or must have so many advantages that you simply cannot refuse.
- Vendors need to agree upon a single standard, or somebody with a large impact should "dictate" it in the worst scenario.

Reading TFA, I was quite disapointed, because anything about how this transition to this cleanslate network seems to be absent at this time. But it is still a research project and maybe somebody did learn something from the IPv6 "fiasco".

Re:What are the odds

[mrchaotica]

The flip side is that some of your suggestions can have detrimental effects too:

- Better support for quality of service from end-to-end.

In other words, better support for introducing favoritism between ISPs and content providers, so that (for example) AT&T can extort money from Google and shut down BitTorrent. No thanks; I prefer the "dumb," route-everything-equally, neutral Internet we have now.

- A much better built-in protection against DDoSes and other kind of abuses.

And much better protection against free speech, anonymity, etc. Again, no thanks.

- Vendors need to agree upon a single standard, or somebody with a large impact should "dictate" it in the worst scenario. [emphasis added]

Yeah, that "somebody" being AT&T or Microsoft, who would undoubtedly screw it up with Treacherous Computing, built-in "micropayment" toll booths, and assorted other bullshit. Still sound like a great idea?

Re:What are the odds

[HappyEngineer]

That's not really a great comparison. IPv6 has no immediate benefits and has some short term problems. Using IPv6 is more like paying off the budget deficit. You don't do it because it's not your problem, it's your kid's problem.

Presumably a non-trivial increase in connection speed would be a much bigger draw to people.

Re:What are the odds

[Bozdune]

Brilliant post.

Re:What are the odds

[Anonymous Coward]

man, what you are looking for is Netsukuku!
it has even much more than what you asked...
it will add to your list:
-anonimity
-no need of dns
-crypted packets
-no need of isp (hell, just check!!)
-traffic can be divided in 2 : that which needs bandwith (p2p for wxample) and that which needs low pings (ssh, games...)

Re:What are the odds

[dattaway]

It does have hope. It has a great business plan. You can bet its protected by a large army of patents. The internet as we enjoy it does not have such restrictions. They want a new landscape where everything is owned in such a way where the new generation 20 years from now could be covered in patents too. Technology can always evolve and be forever be covered in new blankets of patents.

Re:What are the odds

[Anonymous Coward]

I'm slightly confused, you don;t seem to be talking about IPv6 at all.

Lets see - your unique identifiers that you can take with you - IPv6 does that, though I believe its not widely implemented. However, you don't *need* to take your number with you - autoconfiguration takes care of that completely. No, its not like DHCP. This is, incidentally, also a partial solution to the growing router tables.

The other solutions to router issues is simply that the address space is so large it can be partitioned much better, so more intelligent routing can be carried out.

Multicasting support is provided by IPv6, not like IPv4 where packets are sent all over the network and each host must listen for them. With IPv6, all multicast packets have a dedicated prefix, so everyone not listening can filter them out at the hardware level.

While you cannot protect against DDoS (after all, is a slashdotting a DDoS, or just popularity), IPv6 does protect a little against it. It also supports IPSec by default, so I guess some of those rules could be used against, say, SYN-style DDoS attacks.

You do not ned to upgrade the entire infrastructure. Its already in place as hardware has supported it for ages. I think its only been the lack of computers with it enabled that have stopped it being used today. That and the fact that IPv4 still works, and genrally users like to stick with stuff that isn't broken. Vista has IPv6 enabled by default, so if an ISP wants to turn on their IPv6 support, they could and users would be happy. If not, you can run over protocols that package the IP up and then restore them, so you can run IPv6 over a IPv4 network. As everyone still has IPv4 stacks, it will still work until you switch to IPv6 - IE and firefox will both use IPv6 if its avilable.

Perhaps you need to read slashdot, a popular technology blog website: http://it.slashdot.org/article.pl?sid=07/03/08/169 206

Re:What are the odds

[MalleusEBHC]

- More flexible routing of unique identifiers (let's call them IP numbers), so I can take my "identifier" with me (think mobile phones)
- A solution to the ever growing "global routing table" (BGP4 as it is used today)

I don't think it's possible to have both at the same time. A solution for a portable unique identifier already exists (DNS), and trying to achieve portability down at layer 3 could get real ugly and computationally expensive. DNS can be distributed very easily and allows leaf nodes to do the URL->IP translation. This leaves the big routers in the middle with a fairly simple routing table rather than needing to have a table entry for each and every IP address.

Re:What are the odds

[griebels2]

In other words, better support for introducing favoritism between ISPs and content providers, so that (for example) AT&T can extort money from Google and shut down BitTorrent. No thanks; I prefer the "dumb," route-everything-equally, neutral Internet we have now. Do you really think the Internet is this "neutral" right now? I've worked for several ISPs and know all about routing traffic the cheapest, yet still acceptable way. In the end, I always was the techie and only wanted to get my traffic to the destination in a way the least users would complain about "speed" without violating traffic commitments from our upstreams. This "net neutrality" is only politically . I'm a big ISP and I want money from Google? I just route all my traffic to Google to this already filled-up-to-the-max transit link and let Google pay for a direct peering with me. The way this works in practice? The ISP's helpdesk will get flooded by complaints and this "upgrade" will be undone within a few days, until the next manager comes by with yet another great idea to make some more money. Being an somewhat honest ISP, better QoS support from end-to-end will give me much more possibilities to deliver services to my customers in a more reliable way. I could, for example, avoid customers line filling up with bitorrent while using Skype. There is no way of doing this right now. So better QoS support across the Internet is really a cornerstone for reliable services delivered across the Internet, especially for a neutral net.

And much better protection against free speech, anonymity, etc. Again, no thanks. In an Internet without any protection against those kinds of attacks, the one with the biggest botnet wins? There are many ways to implement this kind of protection right into the protocol, without losing any kind of anonymity. Detecting and mitigating DDoSes more close to the source for example. Also, when I don't want to receive your traffic, why do I have to block it on the receiving end? How anonymous do you think you really are? Everything you do leaves traces. Posting on slashdot leaves your IP and your IP can always be traced back to your ISP. Your ISP will probably retain some logfiles, like from which DSL line did it come, from which dialup bank, etc. Public WiFi hotspots or some "anonymity services" might give you some anonymity, they will probably also do so in a "DDoS protected" environment.

Yeah, that "somebody" being AT&T or Microsoft, who would undoubtedly screw it up with Treacherous Computing, built-in "micropayment" toll booths, and assorted other bullshit. Still sound like a great idea? Many of the not-so-evil standards we use today were originally conceived by private or public companies. Sometimes you cannot rely on "standards organisations", because they just are so damn slow and have a tendency to come up with standards that are to much of a compromise. Fortunately, not all companies think they can rule the world alone. For the remaining companies, let's hope they see their quasi-monopolies erode in the end.

Re:What are the odds

[mrchaotica]

I could, for example, avoid customers line filling up with bitorrent while using Skype. There is no way of doing this right now.

Sure there is: you just increase the total bandwidth! That's a better solution anyway.

Detecting and mitigating DDoSes more close to the source for example.

DDOSs don't have well-defined "sources;" that's why they're called "distributed."

Re:What are the odds

[griebels2]

Sure there is: you just increase the total bandwidth! That's a better solution anyway. That's only a temporary solution at best. Increasing bandwith also means increasing usage. In the end, there will always be bottlenecks. In a moving market, we will always see demand for more bandwith. But there will never be enough bandwith. So it is better to the available bandwith wisely.

DDOSs don't have well-defined "sources;" that's why they're called "distributed."

Every packet on the Internet has a source. Distributed only means they come from many sources. Most networks today are smart enough to avoid routing spoofed packets. A future "IP protocol" could totally elimitate spoofed packets. I know of no legit use for spoofed packets anyway.

In a DDoS situation, you can often identify the sources quite easily, but filtering them helps you nothing. The packets still fill up your pipe(s). In a "better world", you could block those packets to you, much more closely to the source, eliminating most of the damage.

DDoSes are a serious problem now and will become even more serious in the future. There are quite some people that already lost ther jobs due to some scriptkiddies playing with the latest DDoS tools.

Re:What are the odds

[Klaus_1250]

Being an somewhat honest ISP, better QoS support from end-to-end will give me much more possibilities to deliver services to my customers in a more reliable way. Implementing any (global) system that relies on people/organizations/companies being honest is silly.

Re:What are the odds

[87C751]

How anonymous do you think you really are? Everything you do leaves traces. Posting on slashdot leaves your IP and your IP can always be traced back to your ISP.

Perhaps you're unfamiliar with IP address spoofing? Granted, egress filtering combats this, though edge routers must enforce it. Once past a common back-routing point, spoofed packets are beyond tracing.

Re:What are the odds

[Anonymous Coward]

You're seeing the "mommy! mommy! it's cold when I jump in the pool!" syndrome.

It will pass... like a bullet through your spleen, but it'll pass.

Re:What are the odds

[curunir]

In other words, better support for introducing favoritism between ISPs and content providers, so that (for example) AT&T can extort money from Google and shut down BitTorrent. No thanks; I prefer the "dumb," route-everything-equally, neutral Internet we have now.

There's a huge difference between allowing QoS at the transport layer and the non-neutral routing solutions proposed by the telcos. QoS makes a lot of sense. It enables new internet services like IPTV and helps ensure that VOIP / videoconferencing and the like are reliable.

The key is, that it needs to be neutral to who is sending it, not what is being sent. If the telcos are allowed to extort money from service providers based on not routing their specific traffic as highly as the traffic for someone else, that's going to suck for everyone except the telcos. But if the telcos are allowed to prioritize VOIP traffic ahead of web/email/torrent/etc, that makes sense so long as they prioritize all VOIP traffic.

Re:What are the odds

[Anonymous Coward]

Outside stuff like DOS attacks, good luck doing anything useful on a spoofed address. Most activities require bidirectional traffic, which tends to go bad with spoofed addresses...

Re:What are the odds

[sgtrock]

Unfortunately, IPv6 really fixes none of those problems, except the IP number shortage.

This is so wrong I don't even know how to respond. For someone who claims to be a former "tech at several ISPs", you are sure ignorant about the design goals and achievements of the IPv6 team. Every single item that you claim that IPv6 can't do, it does quite well.

Re:What are the odds

[Anonymous Coward]

There are some 16 gollabazillion IPv6 addresses: a fairly large supply. There is fairly little demand for them.

They cost how many thousands a year for the smallest available block? This clearly does not reflect a healthy market.

As an ISP, I can't justify the expense. It makes no economic sense. Especially considering that if I put my whole network on IPv6, I would still have to NAT everybody to IPv4 so that they could experience the web. Oh, and that's right, there is no IPv6-to-IPv4 NAT technology that would facilitate this... Oh yeah, and there are a lot of my customers who can't run IPv6 on their PCs. But I could tunnel their IPv4 connections across my IPv6 core network to dump them back onto the IPv4 web.

So I could pay thousands of dollars a year for a block of IPv6 addresses that would enable my customers to connect to... nothing? Wonderful.

Re:What are the odds

[Cato]

You are really very wrong about IPv6, and deployments on the ground back this up - the cost of config is real but most core equipment already comes with IPv6 supported, so you just have to enable it.

The largest US cable operator, Comcast, has already enabled IPv6 alongside IPv4 in its core network and has plans to deploy it all the way to the edge, largely because it needs 100 million IP addresses for its customers and has already used up 16 million addresses (the whole of the 10.x address space, it's having to use public IPv4 space already). See http://www.6journal.org/archive/00000265/01/alain- durand.pdf for some details - they will go dual-stack in most parts of network but IPv6 only for new set top boxes (STBs) and other home network kit such as cable modems (CMs) and VoIP MTAs.

I agree that IPv6 doesn't solve all the problems mentioned, but then it was never meant to. Its enablement of end to end services without the innovation-hostile NAT does mean that you have a better chance of clean solutions to problems such as QoS. IPv6 does address Mobile IP much better than IPv4, so you can keep the same IP address as you roam from one mobile network to another, without inefficient triangular routing of all traffic via a home agent.

Re:What are the odds

[Cyberax]

DDoS packets are not 'spoofed'. And they form valid 100% standards-compliant TCP sessions.

How are you going to distinguish between 10000 Slashdot readers hitting a server and 10000 botnet computers.

Won't work IMO

[zappepcs]

No matter how good a set of tools you make, some^H^H^H^H most people will use them incorrectly. I have yet to see a corporate network designed in a way that both makes sense and is secure at any place I've worked or knew anything about, despite all the good information available on how to do both.

Re:Won't work IMO

[jandrese]

Most corporate networks make sense when they were first deployed, but that was back in the 80s and the technology (not to mention corporate layout) has changed enough that it seems crazy today. I know our tech guys here work really hard to keep everything up to date, and for the most part our network is sane, but sometimes there are cases of legacy systems that really look out of place next to everything else.

I want to know how they're going to avoid the second system effect with their new internet. One of the big reasons the Internet works is because a lot of effort was spent in keeping everything reasonably simple. Time has shown that anything that start out highly complicated tends to be only very slowly adopted, if at all. IP may have terrible security but at least it doesn't require someone 10 man-years to build a fully compliant router.

Re:Won't work IMO

[peragrin]

I think it would be far simpler to first build new protocols, to replace things like smtp, and pop first. kill off FTP to replace it completely with SFTP.

Once that part is done moving to better hardware will be easier.

Re:Won't work IMO

[CastrTroy]

But if you're just transfering publicly available files over using anonymous accounts, then what is the point of SFTP? I understand the need to get rid of telnet, which I would assume is never anonymous login, but things like FTP do have their place. Why not just get rid of HTTP, and use HTTPS all the the time?

Re:Won't work IMO

[jandrese]

I think the argument is to use HTTP for all anonymous file transfers. I'm ambivalent on that solution because current HTTP clients don't support resume or directory listing (sometimes you just have a ton of files you want to make publicly accessible with a minimum of fuss) and have no standard way to upload. On the other hand, the way FTP manages the connection (not to mention the confusion over passive vs. non passive) leaves a lot to be desired on the current internet (FTP hates NAT).

Re:Won't work IMO

[peragrin]

FTP is the telenet of file transfer protocols. it transmitts passwords as clear text. HTTP isn't designed as large file transfer protocol, it works great on numerous small files, but you can always get a better connection with ftp, or sftp.

Re:Won't work IMO

[CastrTroy]

But if you're using FTP for anonymous file transfers, then there is no password, some clients send the email address, or a fake one you've typed in as your password. Some servers accept anything as a password, even a blank password. The extra overhead for SFTP may not really be a big issue, but there's no reason why FTP can't be used. There's nothing wrong with it. If you are having user logons, then maybe you should migrate to SFTP. However, I don't know a lot of people using it. I took a networking course, and the professor was talking about how SSL was the wrong solution for the problem. This is because the hackers didn't actually listen over the line, what they really did was break into the databases where the retailers stored credit card numbers and such. What they really needed was a system where the retailer could collect payment without the credit card information ever having to be on the retailers servers. Right now it's up to the retailer to make sure they're handling your credit card information correctly.

Clean Slate Precursor

[Stanistani]

>a clean slate approach to enterprise network security (Ethane).

Kinda flammable, and not shiny enough. I suggest we take it one step further and use ethylene.

I'll get right on that...

[Anonymous Coward]

And once it's completed, will the switch to it be as blindingly fast and as painless as moving to IPv6? Oh, look! An X.25 gateway!

anonymity vs. accountability

[Red Flayer]

Can be found here, is linked to within the first link provided in the summary.

One of the most interesting criteria for a new internet, to me, was criteria #7:

Support anonymity where prudent, and accountability where necessary.

Maybe it's just me, but it seems true anonymity is becoming more and more important, and less and less available, as governments snoop more on the internet.

Re:anonymity vs. accountability

[nine-times]

Yeah, I'm not sure how to fix this, but it seems to me that it's the single greatest problem with the internet. If you really know what you're doing, you can stay anonymous when you want to do something nefarious. However, if you're just a standard know-nothing user, all your innocuous activities are recorded all the time.

That's the exact opposite of what you want. It's not an unusual sort of security problem, and like I said, I don't know how to fix it because how do you distinguish between nefarious and innocuous? Still, it seems to me that it's at the heart of the virus/spam problem. How do you make sure that e-mail is coming from a valid source (which would allow you to eliminate SPAM and e-mail viruses) without requiring everyone to register their e-mail address to a real-life identity (thereby destroying anonymous e-mail).

Re:anonymity vs. accountability

[ScentCone]

Maybe it's just me, but it seems true anonymity is becoming more and more important, and less and less available, as governments snoop more on the internet.

On the other hand, unless you want this to be a tool only for and by the government, you've got to get businesses comfortable with it. Banks. Retailers. Airlines. Anonymity (of the you-can't-track-my-pr0n-use, or the posting-as-a-troll, or the PRC-can't-ID-the-rebel variety) is antithetical to trustworthy transactions, and without money changing hands, the plumbing is WAY less useful to the huge swaths of the economy that would fund (indirectly) the growth and adoption of such a thing.

"Where prudent" and "as necessary" etc., are completely subjective. People who like to rip off movies have one set of priorities, and people who administer your payroll or need to transmit your cancer meds prescription are looking at it from a very different perspective.

Re:anonymity vs. accountability

[nine-times]

Anonymity (of the you-can't-track-my-pr0n-use, or the posting-as-a-troll, or the PRC-can't-ID-the-rebel variety) is antithetical to trustworthy transactions.

But that's not to say that they can't happen over the same infrastructure. Even today, you can send an e-mail with a fake address routed through some random SMTP server and it's pretty hard to trace. -or- You can digitally sign and encrypt e-mail traffic. Assuming the infrastructure can support both, it's a question of whether endpoints will accept both.

Re:anonymity vs. accountability

[beyondkaoru]

digital signatures -- even if you can't tell who is sending some packet, and in the internet protocol we use as well as ipv6 a router can easily change a field, you know who originated the information, because of cryptography. the key thing is that we can have anonymity when we want, and can prove who we are if we want. banks typically require some sort of authentication already (even if it's as insecure as a social security number), and adding anonymity to the underlying protocol would not change that.

Re:anonymity vs. accountability

[rasputin465]

Yeah, I agree. Besides, `anonymity where prudent' isn't really anonymity, is it. That's like saying `you are free as long as you do everything i say'.

WTF is Ethane?

[drinkypoo]

Besides the obvious, I mean? This is what is wrong with using common words as names for major projects. You can't find them with google!

Re:WTF is Ethane?

[DBCubix]

Ethane sounds like 'vapor'ware to me. haha

Re:WTF is Ethane?

[jim3e8]

Generally true, although the highly popular projects such as Gallery tend to rise to the top nonetheless.

Re:WTF is Ethane?

[VitrosChemistryAnaly]

You can't find them with google!

I typed in (or rather, copied and pasted) "enterprise network security (Ethane)" into google. You'll notice those are the last 4 words in the article summary. My first hit was this website.

What's so damned hard about that?

Re:The Plan

[gregleimbeck]

Umm...IE maybe?

ChangeOver

[imscarr]

Now if we can all just leave the planet for a while, while the people in charge can do the changeover...

Re:ChangeOver

[Iphtashu Fitz]

Now if we can all just leave the planet for a while, while the people in charge can do the changeover...

Just wait a few months for the USA to invade Iran, causing both Iran and North Korea to launch nukes, followed by the USA's inevitible retaliation. All the EMP's will burn out the vast majority of the existing internet, paving the way for the quick adoption of all this new technology!

Re:ChangeOver

[Hoi Polloi]

I just look forward to having all my prOn in one place on the net and properly categorized.

Hasn't this been tried before?

[michaelmalak]

I think it was called OS/2. Or maybe 68000. Or was it Itanium?

Re:Hasn't this been tried before?

[nine-times]

Yes, a great many projects that aim to "start from scratch" don't really make it. However, it's often the case that starting from scratch enables people to think about solutions from a fresh perspective, without all their old assumptions. Even if the actual "from scratch" product never really comes about, or if it comes about and is unsuccessful, often the solutions and the fresh insight creep into the old legacy systems' updates.

Re:Hasn't this been tried before?

[LWATCDR]

Actually the 68000 was very successful. It is still found in many embedded systems and sold millions of units.
OS/2 failed not because it was a clean sheet but because it wasn't. IBM insisted that it run on the 286. Microsoft wanted to drop the 286 and design a version that would be multi-platform and 32-bit so IBM pushed ahead with Microsoft's help with OS/2 2.0 and Microsoft started work on Version 3... They later stuck the windows GUI on it and called it Windows NT.

Itanium? Who knows. The PentiumPro as looked down on because it ran 16 bit code slower than a Pentium it was only used in some high end workstations and servers. Later it grew into the Pentium III.

Re:Hasn't this been tried before?

[kad77]

Over a quarter billion 68000 series CPUs (including its direct variants) have been manufactured to date (probably, that particular design is still very active after 20+ years).

It's success/failure is not even remotely comparable to OS/2 or the Itanium... get a clue!

Re:Hasn't this been tried before?

[TheRaven64]

Itanium is an interesting research project, and it may well fit in very nicely with future systems (I have a few JIT systems in mind that would be better suited to Itanium than any other current CPU). Things often don't end up doing what you expect, however. The predecessor to Itanium was the i860, which eventually found its way into a lot of workstation graphics cards (due to its vector processing ability), but failed completely as a general purpose CPU. The m68K was successful as a microcomputer CPU, but probably sold almost as many in the MMU-less Dragonball variant found in early Palms, and definitely sold a lot more to embedded (primarily telecoms) systems developers. You can now get m68K variants that run at over 300MHz. I'd love to have an Amiga laptop built around one...

Re:Hasn't this been tried before?

[LWATCDR]

The M68k is also very popular in automotive applications.
An Amiga laptop? I would rather have a PPC Amiga laptop with a JIT compiler for old code.
Or if you could reprogram the Transmeta chip to run 68k code it could be very interesting.

Re:Hasn't this been tried before?

[mgblst]

Note quiet sure about your point? The 68000 is still one of the most popular chips ever created, and is still being produced 27 years after it was created.

Even discounting that, are you trying to say we shouldn't try to make things better, but some things have failed in the past. Why don't you add to your list the ipod, Mac OSX, Jet airplane, Linux... the list is endless.

Insightful... no, but maybe In-sigh-full (to insight sighing)

What I really want

[Anonymous Coward]

If they make a second Slashdot, I hope it will have a better dupe checker.

Re:What I really want

[MobileTatsu-NJG]

I was just thinking... if they ever did a new version of Slashdot, it should have a better dupe checker.

Who's In Charge?

[adavies42]

Unless this is being run by the IETF with EFF looking over their shoulder the whole time, I don't trust this to end up as something I want to use.

Re:Who's In Charge?

[JohnnyGTO]

Amen to that brother. It would be like the current boobs in office rebuilding the American Constitution!

Re:Who's In Charge?

[kinglink]

I only approve if EFF has a giant muzzle on their mouth. I applaud EFF on a few occasions, however in the past couple years these are few and far between. Basically they are sorta like the child who kicks and screams every time something doesn't go his way, yet at the same time they seem to miss important fights. The 2600 court case they didn't seem interested in, but they are willing to attack anything the government does.

Basically they feel like they are anti-government, more than they are pro-freedom, and while they can sound similar they are not.

Btw this will never work because no one is going allow the internet to be completely recreated from the ground up. Removing analog television and trying to implement IPV6 should have proven that a long time ago. Neither is going to work until they MUST work. And that's years off.

Ethane ?

[Anonymous Coward]

Is it a Green House gas ? I guess that they are working on a protocol named Kyoto ?

Just my two cents.

Clean Slate Design

[giafly]

...sounds so much better than Not Invented Here

This reminds me of Meskimen's Law...

[wuie]

"There's never time to do it right, but always time to do it over."

Re:This reminds me of Meskimen's Law...

[starseeker]

As frustrating as it may seem, there are actually fairly sound reasons for this in some situations. I would argue the internet was one.

In theory, ten years of computer science research might have produced a better design for the internet than the one we have today, back when it was first being developed. However, we have learned a lot from the scale-up that on a practical level would be fairly hard to duplicate in a research setting. Sometimes you just don't think of the possible consequences until you see them happen, particularly things due to human beings TRYING to bring down the system. Think about how long telnet lasted, for example.

In all honesty, it's a miracle the world wide web has scaled the way it has - consider the original scope of the military networks and the small amounts of data they were transmitting. The original designs were to Get Something Working and Justify Our Budget - that's how it has to work. I'd say the return on investment for the various stages of the internet has always more than justified even the costs of redoing it. Sometimes you can't wait to figure out how to do it right, because that will take too much time and what you can build NOW is still useful. Think about automobiles - 10 years from now we will undoubtedly be building better ones than we can build today, but the costs of waiting until we know how to do it "right" are much higher than the costs of replacement.

Now, of course, the question of knowing how to do something right is distinct from doing correctly what we already know how to do - one is a research problem, one is an implementation problem. I'm inclined to think that the web is more of a research limitation than a "do it right" issue, although I could be wrong - it depends on how much was known in the beginning states.

Re:This reminds me of Meskimen's Law...

[jgrahn]

["There's never time to do it right, but always time to do it over."]

As frustrating as it may seem, there are actually fairly sound reasons for this in some situations. I would argue the internet was one.

I think he was mocking the clean slate scheme, rather than criticizing the original design of the internet. As far as I'm concerned, the internet was done right (which doesn't mean it was finished and carved in stone thirty years ago, but rather the opposite).

In all honesty, it's a miracle the world wide web has scaled the way it has - consider the original scope of the military networks and the small amounts of data they were transmitting.

Uh, the WWW is not the internet.

Re:This reminds me of Meskimen's Law...

[Anonymous Coward]

You sound like one of those "incremental developement" advocates. So you spent your youth gaming and haxxoring when you should have been learning about engineering and design.

The internet is based on a handful of up-front designs, which *were* in fact based on extensive computer science research as well as a clear understanding of planning and systems analysis (which you don't learn from world of warcraft, btw). TCP/IP, HTML, SMTP etc were all basically correct and usable at their first iteration. Some fixes have been needed, but most of the changes have just been bells and whistles of dubious value, quality and inter-operability. More like "decremental development" in fact.

Wanna know why the internet, and software in general, is in such a poor state? Because it's being developed decrementally by gamers and haxors and the real engineers don't get a look in any more.

Re:This reminds me of Meskimen's Law...

[Anonymous Coward]

"TCP/IP, HTML, SMTP etc were all basically correct and usable at their first iteration."

Correct and useful, but not final. In particular, the security problems of public networks have loomed large in recent years. And HTML is demonstrably insufficient for the types of content people want to deliver today. Since so many of those content forms were not reasonable or even possible when the original protocols were designed, it is unreasonable to expect them to handle it cleanly. Never the less, holding up implementation to solve what, at the time, would have been extremely abstract and theoretical problems not seen in the wild would have been useless.

I agree that engineering and design are seriously underused in the world of computer software, but the question of what to do and when is almost always driver first by economic motivators and second by technological onces.

Clean Slate vs. Gummed-upTubes

[digitaldc]

"With what we know today, if we were to start again with a clean slate, how would we design a global communications infrastructure"

Get rid of the porn, scam sites and domain squatters - however, this may not be possible.

Re:Clean Slate vs. Gummed-upTubes

[drinkypoo]

Get rid of the porn? That's what the internet is for. Everything else is just interfering with porn.

Re:Clean Slate vs. Gummed-upTubes

[daeg]

What's wrong with porn? The network design shouldn't care about content. That's a place for your personal morals or corporate rules, not network topology.

Re:Clean Slate vs. Gummed-upTubes

[owlnation]

Get rid of the porn, scam sites and domain squatters - however, this may not be possible.

I'm with you on the domain squatters and scam sites. But you'll take the porn from my cold dead... no, wait... warm sticky hands.

Re:Clean Slate vs. Gummed-upTubes

[PitaBred]

Dude, no

(Yes, I know the song is from Avenue Q... it's still a funny video)

Re:Clean Slate vs. Gummed-upTubes

[Joebert]

warm sticky hands.

You sick bastard !

I do it in the shower, I have a great memory.

Thats it... I'm gona make my OWN internet.

[Kenja]

Thats it... I'm gona make my OWN internet. With blackjack, and hookers. In fact, forget about the blackjack and the internet.

Re:Thats it... I'm gona make my OWN internet.

[no_pets]

Hey, I know that you were joking but if some big, redo of DSM, gov't controlled Internet came to be I would certainly hope that ad hoc wireless "internets" would pop up and connect to each other. Or at least some encrypted version would run on top of things.

From overlooked-irony dept

[carpeweb]

Admittedly, this is a quibble and slightly off-topic, but they could use a clean slate for their web design. It doesn't fit in my 1024x768 display.

Re:From overlooked-irony dept

[Anonymous Coward]

I'm not sure it's 100% off-topic...

I'd have more faith in their ability to design an internet if they had spent the 5 minutes it would have taken to proofread their home page.

And with all those corporate monies you'd think they could hire a decent web designer.

Design isn't everything

[Dancindan84]

Often times the best designed technology will lose out to ones that are either marketed more aggressively or are easier to implement. That being the case, inertia is going to be a big factor in this (current internet is already implemented and works fine enough for most people). Something either about the design or "marketing" (government push?) will have to be impressive enough to overcome that inertia. It will be interesting to see if/how that happens.

Re:Design isn't everything

[ivan256]

I think your point is deserving of a mixed metaphor.

This new design is going to get steamrolled by the freight train that is Internet 1.0.

There. Much better.

Re:Design isn't everything

[maxume]

Why exclude ease of implementation when comparing designs?

Re:Design isn't everything

[Dancindan84]

You shouldn't really, but I meant one that is more easily implemented will be chosen over one that functions better.

interesting....

[Anonymous Coward]

It's an interesting idea considering the internet as it exists today was not designed for the types of usage we're seeing. We've bent and patched and hacked it all together so it'd work well enough but the efficiency and security aspects are seriously lacking.

So maybe we should put this out to the /. experts: If you were going to design the Internet today, knowing the kinds of problems we've seen and knowing the type of usage and availablility people expect, how would you implement it? And would you attempt to make use of the bazillion dollars of existing infrastructure hardware or start COMPLETELY from scratch?

Pushing the envelope with scratch

[lawpoop]

For a great history of scratch, check out the documentary Scratch.

I'm been a fan of scratch technology ever since DJ Qbert made the album Wavetwisters completely from scratch, and later made the animated film Wavetwisters from scratch. Now they're making the internet from scratch technology. Which makes sense -- in my mind, scratching is basically analog computation.

Oh yeah, we really need this :(

[Ancient_Hacker]

Hmmm, yep, let's get the experts to redesign the best network ever made.

Let's get the guys that designed all those "wonderful" networks:

• Morse Code
• TeleText
• Telex
• DECNet
• IBM's VTAM
• IBM's CICS
• IBM's SNA
• Banyan Vines
• AppleTalk
• TELENET
• CDCNET
• IBM's LU 6
• ISO net

Oh yeah, let's get the "EXPERTS" involved!

Interesting

[claes]

I would like to see similar a clean slate approach for Unix as well. For example, I am interested in the question - how would Unix work differently if extended attributes were available in all Unix filesystems from the beginning. Tradition often holds back innovation, I feel

Re:Interesting

[peragrin]

Check out plan9. It was created by the same guys who built the orginial Unix to address some of the complaints they had with it.

What I like about plan 9 is that it would work with everything. You could install it on a tv to act just as a remote or local display. it doesn't care.

with plan 9 the network is just another conduit for passing back data. it doesn't matter what physical resource you are using or where on the network it is located. To the OS it is all the same.

Re:Interesting

[EvanED]

For example, I am interested in the question - how would Unix work differently if extended attributes were available in all Unix filesystems from the beginning. Tradition often holds back innovation, I feel

Fully agreed. For instance, NTFS supports alternate data streams, which are essentially really huge extended attributes. (They're a generalized version of HFS's resource and data forks. A number of other filesystems support similar things now too, such as HFS+, ZFS, and ReiserFS4 v4 in a slightly different manner.)

But the problem is that no one uses them because nothing was built to work with them. If you upload a file with the alternate streams, you lose the streams. If you copy a file to a floppy (yeah, I know) or USB drive, you lose the streams. If you dual boot and copy the file to ext3, you lose the streams. If you say 'cat file1 > file2', with the Unix model this is the same as copying a file, but it would lose streams. The same applies for extended attributes, though maybe slightly less. (Like I don't know if copying a file between two ext3 filesystems will lose them or not.)

It's very frustrating, because there are a lot of really neat things that you could envision doing with this sort of metadata, but no one has support for it.

So I've wondered almost the exact same thing myself... if in 1970, someone added extended attributes/streams to Unix, what would it look like today?

(Of course, I also wonder about things like "what would the world be like if water's heat of fusion was a quarter of what it is" brought about by the spring thaw that's in progress...)

Re:Interesting

[amRadioHed]

If you say 'cat file1 > file2', with the Unix model this is the same as copying a file, but it would lose streams.

cat works exactly as it should in that example. If you are using "cat" to copy files then you shouldn't be surprised if it doesn't duplicate the file exactly because that's not what it's for. It would be a mistake to make cat dump file attributes to STDOUT since that would break most correct usages of the command.

That said, I suppose it might be a neat idea if there was something like STDATT in addition to STDOUT and STDERR. That could be a useful way to do what you want.

Re:Interesting

[EvanED]

cat works exactly as it should in that example. If you are using "cat" to copy files then you shouldn't be surprised if it doesn't duplicate the file exactly because that's not what it's for. It would be a mistake to make cat dump file attributes to STDOUT since that would break most correct usages of the command.

Oh, I realize that. I don't think that that specific example is that compelling against the idea, but it's an illustration of the sort of problems that you'd expect to find.

That said, I suppose it might be a neat idea if there was something like STDATT in addition to STDOUT and STDERR. That could be a useful way to do what you want.

I'm actually thinking something like multiple named streams. So if you have a file with two attributes and cat'd it, the contents of the file proper would go across stdout, the contents of file:att1 across a pipe named att1, and file:att2 across a pipe named att2. Then the shell would direct the pipes to attributes of the same name as the pipes.

Now that I'm writing this out, this would work splendidly with another project I've thought about doing, which is writing a shell or program that would give Linux/Unix the same pipe capabilities supported by IBM's CMS (the OS that was written to be used with z/VM). Commands can have multiple input streams/output streams, and you can connect them up in arbitrary ways. For example, I think there is a filter that sends alternating lines of the input stream to one of n different output streams. (Amusingly enough, Googling "IBM CMS" brings up their page on pipelines as the first hit.)

Anonymous 'net?

[Alcimedes]

I hope in their quest for better security they don't get rid of annonimity (sp?).

What good is a network to exchange free thinking and ideas if Big Brother is looking over your shoulder the entire time.

Re:Anonymous 'net?

[dokhebi]

I would guess your are all for an anonymous internet until someone starts slandering you, then you will want "Big Brother" to step in and prosecute the individual while you slander somone else from the safety of your anonymity.

This is how it always works.

As always, just my $0.02 worth.

Re:Anonymous 'net?

[Alcimedes]

Then you would guess wrong. Thanks for playing.

It has nothing to do with slander, and everything to do with not trusting government or those in power. When people in China are getting arrested for speaking out online against their leaders, it makes me nervous.

How long until disagreeing with the party line makes you a terrorist and yields you the same treatment here?

Re:Anonymous 'net?

[NayDizz]

Reminds me of this assclown. Michael Crook, to be specific.

Involve the porn barons!

[adnonsense]

Get the guys (and gals?) with the high multimedia delivery needs in on it from the start - they'll give you more bangs for the buck for both conception and practical trialling of the new system.

Great!

[140Mandak262Jamuna]

I have already patented scratch. So I am in for a huge stream of royalty payments!

make world

[Anonymous Coward]

Isn't rebuilding from scratch the quickest way to get nothing done??

On that note, I would like to rebuild the world from scratch, to make it more secure, reliable, and to eliminate religion. Who's with me?!

Rebuild the Internet

[hackus]

Translation:

Lets rebuild the internet because it uses too much open source software and we are not making enough money. I know! Lets get all the vendors together and rebuild it using proprietary crud so that it is impossible for any of these "open source" guys to make server platforms that are freely available.

Lets kill open standards too, because well....who needs those IETF guys anyway! They are just a bunch hippies!

Seriously, though. The internet works better than my cell phone does.

It doesn't need "fixing".

It just needs a few upgrades.

IPV6 would be a nice place to start!

GAD.

The thought of CISCO having a hand in anything the future internet could be makes me want to quit my current network manager job and open an Italian Restraunt.

-gc

-hack

Re:Rebuild the Internet

[catbutt]

It doesn't need "fixing". Are you suggesting that the current design is going to be just fine in the year, oh, say, 2500? That we should just say "its as good as it will ever be, for as long as humanity exists on planet earth"?

Progress happens. Deal with it. You don't have to participate if you don't want. Seriously, don't worry, if we need you, we'll call you.

Re:Rebuild the Internet

[Xzzy]

What's to say the internet they create in 2007 will be any more suitable for the year 2500 than what was created 30 years ago?

The point is, if it ain't broke, don't fix it. The issues with the existing structure have already been addressed (IPv6, regardless of adoption rate), so I don't see what advantage there is to further development when we don't even have an idea yet what needs to be fixed.

Re:Rebuild the Internet

[Jeffrey Baker]

I'm with you. These guys are completely on crack. Haven't they ever read "Netheads vs. Bellheads"? You do not want to have intelligence inside the network, ever. Intelligence belongs at the edge. The core should be application-unaware, stupid, unreliable, and as simple as possible. Which is the Internet we have today, and it works great, thank you very much.

Re:Rebuild the Internet

[Anonymous Coward]

...we don't even have an idea yet what needs to be fixed.

OH RLY?

IP spoofing. And multicast is too damn hard.

Whatever happened to the mbone?

Re:Rebuild the Internet

[amchugh]

I thought Cisco researchers contributed something like a quarter to a third of the Internet RFCs?

Re:Rebuild the Internet

[hackus]

I am not suggesting the current design will be OK in 2500.

Furthermore, that arguement is stupid because the people discussed in the story line who propose "fixing" the internet certainly don't have what it takes topside to come up with a design that will last till 2500 either.

Finally this isn't about progress it is about the motive OPERANDI of the people in the story.

Speaking from experience, I just kicked out all of my CISCO routers from my network last year and replaced them with a Linux Kernel and Quagga (bgp) GPL software.

Good riddens I say!

Overpriced junk and slow as crap.

Last year I did a full source code audit of all 15 of my router/vpn/firewall images.

Can you say you did the same?

I hope a CISCO employee doesn't get disgruntled someday and decides to insert a backdoor in that softpak update!

Good Luck with that.

-Hack

Re:Rebuild the Internet

[hackus]

That could or might be true.

But the most important contributing protocols involving networking looks like they come from IBM Research labs.

http://tools.ietf.org/html/rfc1655

Do you have any examples?

-Hack

Re:Rebuild the Internet

[theCoder]

The core should be application-unaware, stupid, unreliable, and as simple as possible.

Well, maybe not that one :)

Re:Rebuild the Internet

[Doctor Faustus]

I'm pretty sure that wasn't a typo. The Internet itself is unreliable, and that's good because the overhead to make it reliable is excessive as long as it works a large majority of the time.

If I have a critical application, and I knew that the design of the Internet would guarantee my packets would go successfully from a router in New York to a router in California and back, I would still put in an acknowledgement system at a higher level to make sure everything was getting from *my computer* to the one on the other end. That makes the lower-level acknowledgement packets superfluous.

Since reliable communication is a common requirement, we have TCP. A reliable equivalent to UDP would be nice, though; I suspect it would see a lot of use for web services.

Re:Rebuild the Internet

[theCoder]

Thanks for posting this because it's a good point. The Internet core isn't reliable in that it it's not 100% reliable. Hosts on the edges cannot assume that packets will get through. So, we write good libraries to do acknowledgments and retries (the TCP of TDP/IP).

Of course, if the core of the Internet routinely dropped lots of packets (i.e., it was really unreliable), most of us would be pretty unhappy. Sure, we'd probably eventually get our data, but if only 50% of your packets got through, your transfers would grid almost to a halt. Exponential backoff on TCP window sizes really slow things down when you start dropping packets.

I don't know what percentage of reliability is required for smooth Internet operation, but I'd guess it's probably in the high 90s. So, while it's true that the design of the Internet is that it is not reliable, it does need to be mostly reliable to work well.

Re:Rebuild the Internet

[Doctor Faustus]

Yes, when your error rates go up, it starts to make sense to put lower-level checks in again.

I seem to recall that WiFi drops around 30% of its packets, so it has its own acknowledgement and retry system that will be running underneath, and invisible to, IP. However, I took networking two years ago and I'm a programmer not a network admin, so don't depend on that being right.

Awesome

[stratjakt]

Committee designed systems never have faults!

Typo

[Anonymous Coward]

The "internets" will no longer be a typo, huray!

Also, check those URLs!

Yeah, good luck with that.

[phillymjs]

I'm sure the RIAA and MPAA won't try to force some kind of low-level piracy-monitoring/reporting mechanism into it. No, not at all.

I see the New Internet joining New Coke in the dustbin of history.

To-DO list

[Xymor]

1- do NOT built-in DRM.

2- do built-in better anonymity and security support.

Re:To-DO list

[Anonymous Coward]

Absolutely. I want to be able to send death threats to anyone, untracibly.

I also need to be able to flood any noob's email with messages they can't stop anonymously.

Then, how about making sure I can grab all the bandwidth I want from my school.

Interconnection?

[Anonymous Coward]

This is all cool, if it'll only work for the 200 people involved. I don't think they should use the term "internet" simply because they think globally. In fact, I think they should call it "alternet", because "the" internet is already here, and we have it. Unless they're interconnected, it's not going to gain adoption save for communities with common goals or practices (inter-university networks, interconnected company branches, etc). And IF they're interconnected as to lure the users of the internet to the alternative, how could the alternet maintain the qualities they strive for (accountabilty, safety, throughput) ? Sounds like another elitist innitiative, and from my perspective it has value as a Computer Science project. NOT as an everyday alternative, because it's just not feasible for the current users to switch.

Not just Stanford

[beaverbrother]

This kind of research isn't just occurring at Stanford. The NSF has had a big push recently to grant this kind of research across the country.

hey, lets revive DECnet Phase V!

[TheGratefulNet]

or, rather, no, lets not.

(and it got about as much attention as ipv6. they both planned for 'big networks' but we all know how popular OSI is, in the real world...)

So they're really gonna swing it...

[blindd0t]

They're gonna download the Internet?

mommy to researchers...

[WheresMyDingo]

"i turn my back for one second, and you go and start scheming again about building the internet from scratch! isn't this one good enough for you? bad researchers! bad bad researchers!"

fork?

[WheresMyDingo]

why not just fork the internet instead?

Gore

[Anonymous Coward]

this is all just a conspiracy to discredit Al Gore

About your sig

[Anonymous Coward]

Got Geometrodynamics? Awe, too hard to figure out? Too bad.

I thought it was about car accidents that involved a Geo Metro

Re:About your sig

[hackus]

LOL

I don't want The Six Million Dollar 'Net.

[Anonymous Coward]

Every time I download a file or visit a web site, I'll have to hear that cheezy "na-na-na-na-na" sound effect.

Re: I don't want The Six Million Dollar 'Net.

[Anonymous Coward]

I wish I had mod points, 'cuz that made me laugh.

They'll end up with a nice neat network

[Colin Smith]

Which doesn't talk to anything.

If it's going to be useful, it has to talk to everything, that's the whole point of the network effect.

Nice try, but you are a little late

[jbossvi]

I would put the odds of this getting implemented at practically nil. If you do not fundamentally redesign most/all of the protocols, you are just refining IPv4/IPv6 to suit your needs. And if in fact you did come up with a "from scratch" design you have the following hurdles to meet:
-port all known software/libs to use the new protocols
-get all vendors of networking equip to issue major firmware upgrades to switches/hubs/firewalls anything that speaks on the network.
-rewrite networking code for top 6 most popular OS's.
-finally port IOS, JunOS, on all the last hardware models of the last 10 years.

then you might be ready to actually implement something, that is of course if you can then talk a good percentage of the planets ISP/Corp/home users to actually upgrade everything for you.

Case in Point: IPv6
It has been around for a decade. it has been ported and deployed onto most major platforms. There is even app and NAT translators on the routers to ease you into it. There is a well known and defined migration path. The US Govt has mandated migration to IPv6 by 2009 (I think).

And you *still* cant get people/corps to start the migration.

We already have a internet, small incremental changes (MPLS,IPv6) are barely tolerated as long as its super easy and you have a big gain.

start from scratch? you are a little late for that.

Design by comittee

[da]

Isn't one of the strengths of TCP/IP that it was designed by a small group of people investigating a problem and who came up with something that 'just works'(tm). I'd be tempted to bet that putting a big committee on the case will kill the project stone dead. Just a thought...

These days ...

[Ungrounded Lightning]

Is someone going to call Al Gore and get his opinion on this?

These days all he'd be interested in is how much power it consumed.

Don't worry -

[wsanders]

- the Internet, as you know it, is already dead.

Re:Don't worry -

[ectal]

What? Did someone unplug it again?

Comment removed

[account_deleted]

Comment removed based on user account deletion

NETSUKUKU!!!

[Anonymous Coward]

you already have one great rebuild from scratch:Netsukuku

just read all the specs.
the only thig it need to be complete is something that permits police to track down malicious hackers. it's too anonimous for now.

Re:NETSUKUKU!!!

[Anonymous Coward]

right. netsukuku is a great project.
here you have the features
feature_list

I'll wait

[slavelayer]

I'll wait for the first Service Pack thank you very much.

Worse than you think

[thegameiam]

Even worse - they'll go straight to IPv6 (because everyone knows you need 2x the Internet to get full-duplex...)

Not now...

[athloi]

But we've finally perfected all the trolls, griefers, basement dwellers, spammers, curmudgeons and porno freaks that a growing network needs!

Re:Question for Slashdot

[Anonymous Coward]

Why bother posting anonymously if you're going to have your username in the screenshot, eldavojohn.

Incredibly naive?

[cyberianpan]

Further, we believe the Internet.s shortcomings will not be resolved by the conventional incremental and .backward-compatible. style of academic and industrial networking research Ok if they were modelling a theoretical utopia & a promised phase2 was then a series of iterations to look at what might be practically achieved but they have the conceit of presenting a viable green field solution ?

Fact is that the existing internet is massive & anarchic - I can't see any viable big bang transition model & parallel running would be fantasy. We've known for decades that the Dvorak keyboard beats the Qwerty but because of transition cost it's not a runner. Ipv4 vs Ipv6 is a struggle even...

For a transformational change project to suceed it would have to be many, many X better than the as-is & this benefit would have to be readily realisable otherwise buy in would be nada.

I'dve no objection if this was a thought experiment/deep research but they are couching it as though it is practical or realistic, methinks they need to add some clinical psycholgists to their staff.

Internet Mail 2000 anyone?

[Inmatarian]

http://en.wikipedia.org/wiki/Internet_Mail_2000

The name is crappy, but the concept is a really good start. It's a shame this never caught on. Basically, Email's Subjects and Bodies are split, and the Subject is sent to the Receiver, and the Body is stored at the Sender's server. When the Receiver gets the Subject notification, they connect to the Sender's server and download the Body.

The point of this strange scheme would be to crush spammers under the weight of their own To list, by having millions of incoming connections. The burden of storage goes to the Sender, not the Receiver.

That should be one of the technologies Web 11.0 should implement. Somebody call up Al Gore and tell him this.

Re:Internet Mail 2000 anyone?

[Anonymous Coward]

Except that botnets would easily handle the load (whereas legitimate sites running huge mailing lists would find it more difficult). Additionally, a subject-only email could easily fool a person (or spam filter) into downloading a message, which automatically verifies for the spammer that the email address is active.

Re:Internet Mail 2000 anyone?

[bcrowell]

I agree completely. However, what this article is talking about is redesigning the lowest-level workings of the internet and its protocols, not relatively high level stuff like e-mail. IMO what's really broken is the high-level protocols, e-mail in particular. Another thing that, with hindsight, is clearly a mistake is http, XMLHttpRequest, and all that; it's clear now that many people want to be able to run something like a GUI application through something like a web browser, but the protocols were never designed properly to allow that. Rather than putting a bag on the side of http to allow ajax apps, the right thing to do would have been to leave http alone, and create a completely different application and protocol that would do what people are trying, painfully, to do with browsers and http.

Another problem is the creep of proprietary formats for audio and video. Mp3 is still heavily patent-encumbered, and the licensing terms do not make it legal for a Linux distro, say, to distribute as many copies of an mp3 library as they like. Video is even worse, because the closest thing to an open codec is theora, and theora doesn't work well enough to be practical. What has really turned out to be popular is to wrap videos in flash apps (the way you-tube does), which piles proprietary cruft on top of proprietary cruft.

We have a whole bunch of technologies that do similar things:

• Java applets are free as in everything (now that sun's java implementation is gpl'd), but users hate them because it takes so long to start up a vm. The java applet security model is also too tight for some purposes.
• Ajax is a botch. It's way too hard to get an ajax app to work on all browsers, in a way that's consistent with what people expect from a GUI app. For example, where I work, we have a new ajax app we're required to use for filing certain paperwork, and it doesn't allow cut and paste. The solution that's been proposed is that we print out the old documents on paper, and send them to a summer intern, who will keyboard them.
• Flash is theoretically open in many ways, but in reality it depends on way too many patent-encumbered or license-encumbered pieces to be appropriate for OSS.

Re:Internet Mail 2000 anyone?

[Inmatarian]

The google TechTalk on this and StubMail implementation actually doesn't even use the Subject, they use a UDP notification packet, which only contains the sender's id and a retrieve URI. Their solution is to say that only packets coming from known contacts (address book and public/provate keyed) should be automatically downloaded.

Re:Internet Mail 2000 anyone?

[SurturZ]

I agree completely. However, what this article is talking about is redesigning the lowest-level workings of the internet and its protocols, not relatively high level stuff like e-mail. IMO what's really broken is the high-level protocols, e-mail in particular. Another thing that, with hindsight, is clearly a mistake is http, XMLHttpRequest, and all that; it's clear now that many people want to be able to run something like a GUI application through something like a web browser, but the protocols were never designed properly to allow that. Rather than putting a bag on the side of http to allow ajax apps, the right thing to do would have been to leave http alone, and create a completely different application and protocol that would do what people are trying, painfully, to do with browsers and http.

Oh thank God I thought I was the only one who thought this. HTML has been perverted to the point where people are designing websites that do the EXACT OPPOSITE of what HTML was intended to do. HTML was meant to deliver (principally textual) content, and allow the client to decide how the content was displayed. Instead we have the whole web design industry trying to make their website render the same on all target devices.

What we need is a sensible GUI-delivery protocol, not filling HTML full of crap. Who the hell thought it was sensible to put executing code into the comment tags in the first place? I blame them, 'coz a hack like that is a great big red flag.

mod parent up.

Re:Internet Mail 2000 anyone?

[Doctor Faustus]

What we need is a sensible GUI-delivery protocol, not filling HTML full of crap.
HTML was pretty much just appropriate for Lynx. Mosaic would have been better off with PostScript.

If they build it, who will come?

[Anonymous Coward]

If they build it, who will come? If it becomes a haven for infringing file sharing, I'm sure the MPAA and RIAA will show up. If anyone shows up, that will bring the spammers (unless they figure out how to make spam not work).

Re:Oh yeah, we really need this :(

[cowscows]

What's so bad about Morse Code? Considering the technology and equipment that it was generally used on, it seems quite effective to me. Just because communications has moved passed it doesn't mean that it was bad.

correct me if I'm wrong...

[Micklewhite]

Wasn't it back in the mid/late 90's folks were saying the internet was just gonna implode in on itself? Whatever happened to that?

new internet

[Sloppy]

Sounds fine, as long as you can tunnel it over IP.

Content Management

[architimmy]

How much of this effort do you think is oriented around builind content managment and DRM like tools into the internet at the foundation. I say leave it as it is. If people need something better let them build it for themselves. The internet just isn't that broken that it couldn't be fixed by simple things like... browsers conforming to standards etc. When you get into all this talk about multimedia content delivery etc, that's just something you build new networks for which layer funtionality on top of the internet in a way that's invisible to end users. Any effort to rethink the way the internet works has more potential to add even more problems than to fix anything.

Maybe I'm cynical but...

[Anonymous Coward]

Why do I suspect this redesign is an attempt to answer the question: "how do we put the information genie back in the bottle?"

so the real question is...

[OiToTheWorld]

How will this help me look at boobies more efficiently?

Re:Oh yeah, we really need this :(

[westlake]

Let's get the guys that designed all those "wonderful" networks:

Morse Code. In general use 1844-1999.
Trivially easy to adapt to almost any form of signaling, including assistive technology for the disabled.

TeleText 1970-to date.
In the U.S. most easily recognizable as Closed Captioning for the Hearing Impaired. But it's the root of the web page and any form of interactive television.

Telex ca 1935-to date.
Rugged, reliable and cheap. In Germany alone, more than 400,000 telex lines remain in daily operation. Over most of the world, more than three million telex lines remain in use. Telex

I could go on, but you should get the general idea.

Not exactly

[mengel]

I couldn't help chuckling as I read the above post, as it outlines all of the things that were presented as benefits of moving to IPv6 when it was initially released. For example:

• There are several mechanisms for running IPv4 and IPv6 side by side, and that was a major part of the discussion in the IPv6 rollout early on. Medium sized chunks of the net were running IPv6 for quite a while, and were routed in and out of fairly seamlessly. transition mechanisms were designed, long before IPv6 was adopted by the IETF. (the linked RFC is from 1995).
• IPv6 designers also put in tools designed to provide for mobile endpoints, although better designs have come out since.
• IPv6 provides and uses multicast addresses as part of it's initial design, and its multicast is being used successfully.

You can claim that the implementations provided weren't good enough (although I'd like to see some actual data to back that up), but in fact the folks that did IPv6 did have all of those goals in mind when they put IPv6 together.

Yes, it's called progress.

[Anonymous Coward]

And before Linux there was Unix; what's your point?

Yeah, the Internet seems pretty great now and all of those products more-or-less flopped, but if nobody ever tried to do something different or improve on something we would all still be running around naked in the woods clubbing our dinners to death. We learn from our mistakes (hindsight is 20/20), and with the knowledge we have today if it were possible to "redo" the Internet I am positive noticable improvements would be made (kiss SMTP goodbye). You don't throw away an idea just because "it's too hard". And it's not as if the Internet is a fixed technology that has always existed in its current form.

Rebuilt OSes/programming languages first.

[master_p]

Current operating systems and programming languages suck big time...fix them, and the network will be easily fixed.

This has been needed for awhile

[GilbertZ]

While I have no idea if these guys will do it right, and lots of people seem to think it will be corporation-oriented, this has been needed for a long time. If they can simply address spam and spoofing IP Addresses, which should be easy to take care of with a redesign from scratch, we would already be so much better off.

A "Clean Slate" approach to the internet...

[Anonymous Coward]

...would be like a "Clean Slate" approach to the interstate highway system. Sure, mistakes were made, some of which cause major problems. It's interesting to see what we can acknowledge in retrospect should have done differently. Some of it will suggest good future directions for upgrades and new features.

But at the end of the day, the current infrastructure is built out, for better or worse. There are too many dollars and too much labor invested that would be essentially impossible to throw away at this point. We built it the way we built it.

Doesn't mean it's no valuable to look at this sort of thing, but this is an intellectual "what if" exercise. Nothing more.

A new Internet economics. A worse Internet.

[Nicopa]

The current internet is to equalitary for them. In their whitepaper they state:

[...] A related issue is that the current Internet does not provide support for differentiating between different packets on economic grounds. For example, two packets with the same origin and destination will typically be routed on the same path through the network, even if the packets have very different values.

"Outrageous! The rich treated the same as the poor!" They want an internet in which a porn movie downloaded by a CEO preempts and disturbs a critical communication from a hospital to an investigation center.

The internet as we have it is an open field. A dumb, simple, protocol so that people can innovate in the sides. This enabled us to be independent from ISP and to design new protocols (Gnutella, Bittorrent, etc.). Of course, they now say that this "dumbness" produced lack of innovation:

Resistance to change is compounded by the end-to-end design philosophy that makes the Internet "smart" at the edges and "dumb" in the middle. While a dumb infrastructure led to rapid growth, it doesn't have the flexibility or intelligence to allow new ideas to be tested and deployed. There are many examples of how the dumbness of the network has led to ossification, such as the long time it took to deploy IPv6, multicast, and the very limited deployment of differentiated qualities of service. Deploying these well-known ideas has been hard enough; deploying radically new architectures is unthinkable today.

It's not clear to me how having a more complex internet in the middle will be able to ease its growth. It seems as the opposite, as more complex middleware will be more complex to upgrade and setup. In fact, the main reason the current internet has "ossificated" *is* dumbness in the middle, but other kind of dumbness. The commercial companies' dumb administrators, dumb managers, who didn't care to provide us multicast, IPv6, mobile ip, IPsec, etc.

The Internet as we have it could never had happened if it were for the private sector. It's too open, private companies don't like standards. See how the classical internet infrastructure got frozen when the commercial companies took over internet in the last century. HTTP, IMAP, POP, HTML, etc. got stuck in their last versions. It's because Internet needs a strong *public* presence. Companies can exist, provide service, but Internet needs a strong presence by the people (in the form of the state..? Universities? I don't know...)

This group is not aiming at a better, utopic, internet. They are trying to recapture what they've lost when their CCITT (X.25, X.400, X.500) network wreck.

Re:A new Internet economics. A worse Internet.

[zippthorne]

Wait.. companies hate standards, yet "froze" them? Isn't that just another way of saying they standardized the standards, which up until that point apparently weren't?

Companies don't hate standards. They hate costs. If you want them to change, you have to convince them that either the increased revenue will offset the capital expense or that the reduced marginal cost will offset it.

Re:A new Internet economics. A worse Internet.

[markov_chain]

I'm puzzled by the famous computer science and networking names on the list of researchers. Maybe it's one of those umbrella efforts designed to cover each researcher's pet project, while getting lots of funding from industry.

Re:A new Internet economics. A worse Internet.

[Nicopa]

Standards are supposed to evolve. Have we already got the best email system that can be designed? The best HTML? Besides, new services need new standards.

Companies don't create standards. They only do that when they are forced. Remember the days before the internet, all those products for building an office's lan... there were no standards there, just products. Companies build products, and create proprietary protocol ad-hoc for those products.

If email would have been invented in this commercial Internet era.. we wouldn't have RFC-822. Just as ICQ inventors (Mirabilis) didn't create a standard for Internet presence.

Here ya go!

[Anonymous Coward]

    H H
H-C-C-H
    H H

Outdoors vs. Indoors.

[Anonymous Coward]

"Maybe it's just me, but it seems true anonymity is becoming more and more important, and less and less available, as governments snoop more on the internet."

The Internet unless explicitly used (SSL,VPN,etc), is an open system. Complaining about governments while ignoring all the other witnesses to your actions is shortsighted (much like the great outdoors is an open system. I see you!).

Re:Outdoors vs. Indoors.

[Red Flayer]

Complaining about governments while ignoring all the other witnesses to your actions is shortsighted (much like the great outdoors is an open system. I see you!).

For me, though, freedom from government oppression is far more important than whether joe blow or $COMPANY can view my activity. Sure, data privacy (with special concerns for identity theft, etc) is important, but I don't think there's any single privacy issue more important than freedom of speech, which depends on anonymity when oppressive governments forbid it.

I recommend...

[Digital Vomit]

I recommend abandoning the current internet to the corporate interests that have defiled it and making a brand new internet for the *people* of the world. No corporations or government interference allowed.

Yes, I know it's a "pipe dream" (or, should I say "tube dream"?) but it would be nice, wouldn't it?

Aquinas Protocol

[Anonymous Coward]

I predict that major government contracts will be awarded to Page Industries, with their Aquinas Protocol specification. It will allow us to breathe safe and contront the world's growing terror problem in conjunction with authorities like UNATCO.

Sure, comming right up!

[Anonymous Coward]

While we are at it, why not migrate all existing computer users from QUERKY to Devorak keyboard layouts.

Since we are dreaming I would like a pony too.

A pink one.

With wings.

-- John

Rebuilding? Yes. From scratch? No

[alexfromspace]

When I looked at the title of the article I had a strong surge of hope followed by a suddent concern for job security and visions of decreasing demand for highly skilled professionals. Well, after overviewing the white paper I was feeling completely secure and once again disappointed.

I find most of the propositions as things that need to get done, but overall it looks like just another patch, although a huge one. Majority of it deals with reevaluating design of the physical layer components and their integration, and although grandiose the rest looks like a list of bugs needed to be fixed.

    Seriously, in order to rebuild internet from scratch, most if not all of software dealing with networking would have to be rewritten in order to go from the 5 layer model to the more proper 7 layer model. That would mean pretty much rewriting huge chunks of linux, unix, apache, throwing out billiions of lines of code and eventually seeing a significant decline in the demand for both hardware and software. On the positive note, it might also cripple windoze, dealing it a death blow.

It is nice to see that Stanford is at least considering to reexamine the subject, since we pretty much owe it them for being stuck with 5 layers :), ouch, :)!

Re:Yes yes yes - I want my circuits back

[neutrino38]

Finally !

The original Internet architecture provided a method to transport packets, and has changed little since it was first proposed. It provides a "dumb" connectionless packet- forwarding packet-switched infrastructure, with high functionality at the edge (the so-called "end-to-end principle"). The Internet provides a single, simple lowest-common denominator best-effort packet-switched datagram delivery service (IP), with fixed-size numerical addresses (one per physical network interface). If an application requires a reliable stream service, it can optionally be provided on top of the underlying unreliable service. Adherence to the end-to-end principle has come with two main costs: loss of functionality within the network, and a lack of innovation. Although the Internet has evolved, it has done

Finally, we will be able to have proper VoIP media handling at the core network

In our work, we plan to revisit many of the basic assumptions of network architecture. Here are some topics we will explore: 1. Flows as first-class citizens. One innovation that we believe to be important is the recognition of flows in the network. We believe flows should be treated as (...)

Oh yes again ...

5. Dynamic circuit switching. If the core of the network is to benefit from high capacity all-optical switching, then should we deploy dynamic circuit switching? If so, how?

Classes of service lads ! I want either

• Give me a circuit for all my bandwidth (all my flows)
• Give me a circuit for THIS service (all flows related to one service / media)
• Give me an END TO END circuit for THIS flow (this session)

The trick is: it's not gonna happend soon. Good article nevertheless. Bookmark in my journal.

Re:Oh yeah, we really need this :(

[rickb928]

Dis Vines all you want, but it did stuff Microsoft didn't get right for a decade. NetWare NDS didn't do much more than Vines had already done.

I still use CICS daily. It works. SNA worked too.

Appletalk seemed to suffer more from the problems with localtalk than anything, and if m0r0ns had read the guides, they might not have spliced cables together so often and hozed it up.

Next thing, you're gonna dis Token Ring and ARCnet, both of which were stable in the days when Ethernet couldn't handle a flourescent lamp next to the cable. And both could actually deliver the traffic they were claiming to.

A lot of that old stuff worked damned well. We don't have to back to message switches and uucopy, but lessons back then should be remembered, not learned again.

ps- I'm not in favor of a 'new Internet from scratch'. We have no idea the problems we'll cause with a 'new Internet'. I'm willing to live with the current ones for a while yet.

-rick

So Bush was right?

[Jon Abbott]

...so Bush was right when he referred to "the internets"?

What needs to stay/come in, what needs to go.

[Jorophose]

Needs to go: (To hell) - Domain squatters - Lack of annonymity/beingannonymous - Government interference with what goes on the Internet (Censoring = Bad, arrests = meh) - RIAA/MPAA/etc. and their actions to "clean" the Internet. What needs to come in: - More fiber optics if possible - More gigabit connections if possible - Less of the older, slower, connection cables - More P2P (When used legally) - A system to allow the RIAA/MPAA to send "Cease and Decist" letters to copyrighted content uploaders - A nice big picture of Goatse with "TAKE IT UP YOURS!" written on it, to be mailed to said terrorrists. So faster connections, more freedom, more goatse. Aaaand, if we're fixing domain names too, so by extension the WWW, implement .xxx, and give free transfers to porn sites to get them on .xxx. Get ISPs and web hosts to not host porn unless it has .xxx. Since the Internet is being updated, this is a perfect time to implement it, since we'll be following them around for a month or so while we move them. If anyone resists, well I guess they should be taken off the web. So if we can move porn to .xxx entirely, and possibly push adult sites (Or those with heavy-vilence/swearing/etc.) there too, we could implement .xxx blockers in net tools. It'll also make looking for porn easier: tell Google to only look for sites that end in .xxx

Re:What needs to stay/come in, what needs to go.

[Jorophose]

Sorry, fucked up the older post.

Needs to go: (To hell)
- Domain squatters
- Lack of annonymity/beingannonymous
- Government interference with what goes on the Internet (Censoring = Bad, arrests = meh)
- RIAA/MPAA/etc. and their actions to "clean" the Internet.

What needs to come in:
- More fiber optics if possible
- More gigabit connections if possible
- Less of the older, slower, connection cables
- More P2P (When used legally)
- A system to allow the RIAA/MPAA to send "Cease and Decist" letters to copyrighted content uploaders
- A nice big picture of Goatse with "TAKE IT UP YOURS!" written on it, to be mailed to said terrorrists.

So faster connections, more freedom, more goatse.

Aaaand, if we're fixing domain names too, so by extension the WWW, implement .xxx, and give free transfers to porn sites to get them on .xxx. Get ISPs and web hosts to not host porn unless it has .xxx. Since the Internet is being updated, this is a perfect time to implement it, since we'll be following them around for a month or so while we move them. If anyone resists, well I guess they should be taken off the web. So if we can move porn to .xxx entirely, and possibly push adult sites (Or those with heavy-vilence/swearing/etc.) there too, we could implement .xxx blockers in net tools. It'll also make looking for porn easier: tell Google to only look for sites that end in .xxx

It can be a good thing...

[TropicalCoder]

I found the concept of rebuilding the internet from scratch quite exciting. Now that we have some thirty years of experience with the old one, what a difference we could make with a new one, while at the same time having a much better understanding of how to build a network that will sustain continuing evolution on into the future.

There are a few essential things missing from the Stanford proposal. I didn't see anything to suggest that they are looking for this to be a truly international collaboration. If it isn't, that would be a very short sited omission. Also needed are the inclusion of social scientists capable of making some value judgments and decisions about how the proposed new internet can encourage social inclusion and break down the digital divide, and political scientists who can suggest how the proposed new internet can enhance democracy and international harmony.

Obviously, as the article stated, there will be resistance from current stakeholders who depend on the internet remaining as it is. Advocates of net neutrality are obviously very concerned, but it doesn't have to be the way they imagine. Imagine every packet has fields in the header that indicate its particular needs, whether that is for guaranteed delivery latency, or low jitter, or priority level, (even varying packet sizes may be useful) and every packet priced. Those of you who download entire movies via BitTorrent will be able to save money by just dropping the packet delivery priority. Really, if you want a certain movie, usually it doesn't matter if you get it today or tomorrow or next week. Imagine if you could set the priority - and the corresponding price per packet so low that it takes a whole week to deliver, but costs you only pennies?

The thing is - the current internet IS broken. The article states that current economics can't sustain it as it is, without going into much detail. They do state as evidence, however, that six out of the seven biggest ISPs have had to restructure in an attempt to sustain profitability. Our society (and more to the point, our economies) are growing more dependent on the internet day by day, but we dare not depend on it as we do. In its current state, it is just too vulnerable. It seems quite possible that some country could declare war and launch endless DOS and other attacks to such a degree that it could cripple our economy.

Imagine if our telephones worked the way the internet works now. Over 90% of all the phone calls we receive would be somebody trying to sell us something. We would be getting calls from people in Nigeria asking our help in reclaiming fortunes. When we call our bank, we may actually end up talking to a phisherman trying to steal our money without realizing it. There would be periods when we simply couldn't call out because of endless incoming calls in a denial of service attempt. I am sure many readers could take this analogy a long ways, but I have made my point. In my opinion, only good can come from the Stanford research if they open to broader input.

Re:Oh yeah, we really need this :(

[LarsG]

I could go on

Please do, you were about to get to the interesting part of the list. :-)

The battle of conflation was lost in 2004

[87C751]

Uh, the WWW is not the internet.

Unfortunately, not everyone agrees on that point. I even emailed David Weinberger regarding his statements, and he replied that he intentionally conflates the internet and web because "that's what the "mass market" has done." (inner quotes his)

when? well after the super volcano

[theBunkinator]

I find it rather amusing that "plans to rebuild the internet from scratch" is followed by a story about the Yellowstone Supervulcano rumbling loudly (http://science.slashdot.org/science/07/03/15/1836 223.shtml). Good timing.

Excerpt from an internal presentation

[Jesus_666]

"Ladies and gentlemen, we present to you our design for the new internet. Its main feature is that it's incompatible with anything we have had before, requiring everyone to replace all of their hard- and software. Also, it's expensive and proprietary, which makes it even more expensive. By ensuring that vital components are covered by extremely broad patents we have managed to make it even more expensive, up to the point of making it economically infeasible for most small countries. Finally, we made it so fundamentally different that everyone doing anything remotely network-related will have to be retrained from scratch, adding even more to an already gargantuan pile of costs and expenses.

Everyone, I think this one's a keeper."

anonymity vs. accountability of SERVERS

[mosel-saar-ruwer]

You guys are all worried about anonymity and accountability [usually referred to simply as "authentication"] from the server's point of view, as the server looks at the various clients it interacts with.

The flip side of all of this is authentication from the client's point of view when the client looks at all the servers it might potentially choose to interact with.

If you're a client without an IP address, and you send out a DHCP packet, how do you "know" that the DHCP replies you receive are genuine, and not phishes?

If you're a client that hasn't cached every possible name resolution on the planet, and if you send out a DNS query, how do you "know" that the DNS replies you receive are not phishious replies directing you to phishious servers?

As far as I know, there are no "trusted" mathematical algorithms for authentication that do not require a central repository of known authentication/encryption keys [typically very large prime numbers], and a willingness & agreement on everyone's part to refer to that central authority as authoritative [and even there, everyone has to trust that their apparent interactions with that central authority have not themselves been phished].

An algorithm that provided for non-centralized "authentication" [if such an algorithm is even capable of existing] could very well be the mathematical breakthrough of the century.

So I take it you were for unification?

[mrmeval]

I'd prefer a brown coat.

Re:Oh yeah, we really need this :(

[Ancient_Hacker]

Morse Code:

Two most common characters, nice and short: E: . T: - Third most common character, very long: O: ---

TeleText and Closed Captions:

CC: Error protocol: Send each character twice. TeleText: Assume everybody has a 16-color 24x80 screen.

Telex:

USe a 5-bit character code, with a half-dozen different conventions for character sets, sometimes a bell on shift-G, sometimes a quote, sometimes a bunch of funny old weather symbols. Miss a downshift and you get a line of numbers and punctuation instead of text. An infinite number of endof-line conventions. Printers that can slice off your fingers. Sweet.

Note the "2000" in the name - redesign is hard

[billstewart]

Redesigning mail is hard, especially if you want it done in a way that lots of people actually adopt. DJB should know - he redesigned a mail server to work around many of the problems of Sendmail - and I appreciate his frustrations. Any time a spam discussion goes on long enough, people rant about how SMTP needs to be redesigned, but almost none of the ideas I've seen actually fix the spammer problem in ways that can be adopted incrementally by users who need compatibility with the existing email infrastructure, and since you're not going to flash-cut everybody in the world to your system, you need to support incremental growth.

The fundamental problem is that spammers and users are in an arms race, and the spammers keep thinking they can Make Money Fast, so they keep coming up with new attacks on our defenses. (I know that Rule #1 says that "Spammers are Stupid", but never underestimate the power of stupid people in large groups...). For instance, moving the cost of storage to the sender doesn't fix the problem, the way it might have back in 1999, not only because spammers can now use zombie armies to store the data, but because the data doesn't really change from recipient to recipient except for some obfuscation that can be generated on the fly; spammers really don't mind if they only have to "pay" to send their latest stock scam ad banner to the 10,000 people who clicked on the message instead of the 10,000,000 people who they currently ship it to, and it lets them collect the IP addresses of probable suckers in the process. And the only people who care about storage are mailbox service providers; legitimate senders don't care because disk space is nearly free, recipients don't care, for the same reason, and spammers don't care because it's the same banner so it doesn't even take up noticeable space on the zombies they're ripping off to host it.

The protocol redesigns that *have* been adopted widely are things like "use blogs instead of mailing lists" and "use IM instead of email" and "use really large webmail services to send email to the other users of those services". Spammers have adapted by using blogspam and IM spam and captcha-recognizers.

The big obstacles to implementing IPv6

[billstewart]

There are a couple of obstacles to widespread adoption of IPv6; some are easier to overcome, and some are seriously hard.

• Scalability is seriously hard - IPv6 was supposed to fix this, mostly by giving us a clean slate to reassign addresses in some hierarchical quasi-geographical aggregation basis, but that mostly turned out to be wildly optimistic handwaving that didn't reflect the underlying needs of business customers. It's not just that businesses want to have their own portable address space so they can change ISPs without renumbering; you can bully most of them out of that if you have to.

  The big problem is that business customers need multiple homing, so their address space (whether it's provider-assigned or their own portable space) is advertised by multiple ISPs, so that if one ISP connection fails they're not dead in the water. DNS isn't enough to fix this - DNS caching means that you can't instantly activate your other IP address, and ongoing sessions that have already done their DNS lookup don't pay attention to it later. There's an ugly standards-process thing called shim6 that builds a sort of session-layer into the protocols, and some people think they'll be able to get end users to adopt this, but it breaks more things than just implementing IPv6 naturally does.
  

• IPv6's cool features keep getting adapted to IPv4 - one of the biggest reasons for IPv6 was supposed to be better security (IPSEC does this for IPv4), and also things like multicast (works ok on IPv4, but ISPs don't have the economic models to support it widely, and IPv6 doesn't provide any special economic models), QoS (several versions work fine on IPv4 and on the MPLS used by many ISPs to haul IPv4 around, but ISPs are only starting to see economic usefulness, and it tends to burn router CPU if you're not careful). Most of the mobility solutions I've seen have looked like tunnels or NAT, so IPv4 can support them. The one feature that really can't be adapted is the bigger address spaces, and NAT, RFC1918, and CIDR have let us limp along.
  

  
  

• Chicken-and-egg problems with content - until there are a lot of users with IPv6 support in their PCs, browser software, and ISP connections, there won't be a lot of cool web content that isn't also available on IPv4 (being "a cool IPv6 demo/implementation/developer site" doesn't count as cool; it's mass consumerism and widespread business-to-business that pays the bills.) And without cool IPv6-only or IPv6-faster content, consumers and b2b end users aren't going to bother. Bram Cohen might be able to fix this by making BitTorrent do cooler things on IPv6 than on IPv4, but basically it's a slow-growth problem. Cellphone companies in Japan and China that do the walled-garden content thing are the only other model I've seen that's got any major leverage, at least until the number of home-broadband users in Asia who aren't behind national-censorship firewalls becomes a lot larger.
  

  
  

• Router performance - High-end routers are expensive, and ISP backbones and Colo center routers need really high-end routers to keep up. IPv6 addresses are bigger, so you can't just use your IPv4 router, which implements most of the high-speed packet dumb forwarding in ASICs and does more complex functions in CPUs - either you get a router with IPv6 ASIC support, or you do most of the work in the CPU. Also, routing tables (see previous scalability rant) get a lot bigger when you use IPv6 addresss, so the routing protocol calculations need 4x as much RAM and faster CPUs. This part can be partly solved by adding Moore's Law, shaking well, and waiting 2-3 years, but of course the amount of traffic on the Internet keeps growing. Some of this is helped by moving more of the backbone onto switched environments (whether Ethernet or MPLS flavored), which lets you move more of the routing work out to the edges, but you still need the horsepower.