Web Running out of IP Addresses

[Harlowe]

I am imagining the collective primal scream of millions of WoW and Habbo Hotel players unable to connect...

http://technology.timesonline.co.uk/tol ... 819803.ece

[Freecare Spiritwise]

IPv6 solves that, and it's even built into Vista, but ISPs aren't adopting it. If it weren't for most networks using NAT, we would have probably run out of IPv4 addys by now. But the solution is right there and ready to go. Most everything supports IPv6 now - it's just a matter of using it end to end.

[Ddrak]

There's massive blocks of unused IP addresses - especially if you reclaim the A class nets that were given to early companies that now just sit behind a couple of isolation parts of those:



Dd

[Freecare Spiritwise]

Yeah, some of the blocks have been reclaimed but it's still a stopgap. The sooner everyone moves over to IPv6, the sooner my toaster can have it's own IP. 32 bit addressing (4 billion unique addys) probably seemed future-proof back in the 70's. That's why I think we should skip over 48 bit and head straight to 64. It's going to take the world long enough to make the switch to 48 bit, so why fuck around and have to go through this again some day.

And the heart of the matter is really that the intraweb thingy is built on ancient protocols that were never intended to be used how they are being used today. I spend the majority of my days writing web applications built on a protocol (HTTP) that was intended to just link documents between extremely nerdy men decades ago. We need a new HTTP type protocol that's designed from the ground up for interactive applications. And the other protocols we use often like SMTP, DNS, etc are also long in the tooth. We need to revamp the whole lot, but it doesn't happen because there's trillions of lines of code out there hard coded to these protocols, so like the Winchester castle we keep tacking extensions onto these obsolete protocols. SMTP/POP3 are abominations. Any first year comp sci student could design better ones based on how email is used today. The original designers of these protocols were thinking on a large scale, which is why the internet works, but they weren't thinking on a truly massive scale, where 6 billion people might all have 10 uniquely addressible devices. And they were certainly idealists - spam, armies of bot nets, phishing, viruses, etc., were incomprehensible at the time.

So really I think IPv6 will end up being a stopgap too in hindsight. TCP is stale with 16 bit port addressing, so why not re-do TCP/IP from scratch instead of mutating it.

[Klast Brell]

Yeah, some of the blocks have been reclaimed but it's still a stopgap. The sooner everyone moves over to IPv6, the sooner my toaster can have it's own IP. 32 bit addressing (4 billion unique addys) probably seemed future-proof back in the 70's. That's why I think we should skip over 48 bit and head straight to 64. It's going to take the world long enough to make the switch to 48 bit, so why fuck around and have to go through this again some day.

Nerd to Freecare.
IPV6 is 128 bit. Not 48 Bit. That's about 2 to the 98 unique IP's per person on the planet. About 5 followed by 27 zeros.

I think we will be on for a while with IPV6

[Ddrak]

I spend the majority of my days writing web applications built on a protocol (HTTP) that was intended to just link documents between extremely nerdy men decades ago. We need a new HTTP type protocol that's designed from the ground up for interactive applications.

I think you're confusing HTTP with HTML. HTTP really is just request/response, which can form the foundation of pretty much any protocol you want to write.

Dd

[Freecare Spiritwise]

@Klast - oops, my bad as the kids say. I haven't really worked with IPv6. Not that many people have. I was wrongly figuring that the xN was number of octets (4x8=32 and 6x8 = 48). Either way, ISPs should just make the switch. And yeah, 128 bits, should last us 100 years or so lol.

@DDRAK - I should've said that the HTTP/HTML combo is lousy for web applications. Yeah, the apps themselves are markup, but HTTP was created to serve markup documents, not apps. Hyper text transfer protocol. Most everything you do on a web page triggers an HTTP postback - ugly. You can get around that limitation of HTTP with client side scripting, but that's ugly too (even with AJAX). We need something like an event-driven, object-oriented application protocol (OOAP?) Web applications are at least 10 years behind Windows applications IMO. HTTP/HTML has made us de-evolve. So I'll disagree that the requst/response paradigm is an accpetable foundation for applications.

HTTP is even obsolete for serving up documents. It should have compression built in. It should have "server push" (events) built in. It should be able to process requests and serve up documents in binary. Converting binary documents to text, and converting them back to binary on the client is not only ugly but very inefficient. Massive amounts of bandwidth (not to mention CPU clock cycles) are being wasted.

[Klast Brell]

@Freecare - I would not have known myself, but we just covered IPV6 in my class last Wednesday.

I'm feeling cynical today. So I'll look at the glass as half empty. Every time technology advances some other coder comes up with something that completely uses up all the extra new resources.
Computers got faster, So along came Vista. OK maybe Vista is a bad example. How about Crysis, or any other recent game that pushes so may polygons that the fastest computer possible just 4 years ago won't run it. OK so anyway. Someone is going to see all these I addresses going to waste and come up with some brilliant killer app that really is useful. But also hoovers down IP addresses like they were Cheetos. Maybe there is something really cool you can do if you give each email it's own unique IP. And people will say. "Hey we have IP's to burn" And in 2040 some other generation of wise asses will be cursing us for not thinking big enough just like we are bitching about DARPA.

[Embar Angylwrath]

Klast -

What you're seeing is akin to evolutionary pressures. New computer environments emerge, and new technologies and solutions rise to meet the environment. Business is much like biology. If more people understood that, business majors would be required to take biology and ecology courses.

[Ddrak]

@Klast:

We will never use all IPv6 addresses. Seriously. You could give an IP to every particle in the universe since the big bang and STILL have orders of magnitude left over. 2^128 is a seriously large number.

The problem isn't the IP addresses, it's the routing tables. How you know where to send this packet you have for the IPv6 address you need to get it to. Core routers are already memorizing the entire IPv4 space and IPv6 just makes it worse - that's why you haven't seen a fast transition.


@Freecare:

HTTP was created as a general purpose request/response mechanism. It doesn't mandate the content as being text - you can send whatever you like after the header. In fact, this web page was served up as a binary .gz compressed document. Similarly the "postback" thing has nothing to do with HTTP itself - it's the browsers that enforce that. I agree that not having a persistent connection to allow server push is somewhat troubling though. Maybe HTTP 1.2?

Web apps are never going to be as good as thick-client apps. You just don't have the bandwidth to do everything you want.

Dd

[Freecare Spiritwise]

I was creating TCP/IP based, n-tier client/server Windows applications in 1994 under Windows NT 3.1. In many ways they were superior to what I'm creating today. The problem is that our clients, and pretty much the whole world is demanding thin-client web applications. I get it - all you need is a browser - there's no installation/compatibility/upgrade issues. All you need to use a web app is a computer and a URL. It's convenient, but from a development standpoint it's a bloody nightmare by comparison. The development time (i.e. cost) is about 3 times longer and you're getting less functionality than its Windows counterpart. But you gotsta give the people what they want /shrug.

[Ddrak]

Yep. Agreed on all counts there (and I was doing the same thing in '94).

Dd

[dorick]

Here's a quick visual.

IPv4 (32 bits)
00000000.00000000.00000000.00000000
Possible Addresses: 2^32 = 4,294,967,296


IPv6 (128 bits)
00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000.00000000
Possible Addresses: 2^128 = 3,402,823,670,000,000,000,000,000,000,000,000,000,000 Yeah, that's a lot.