IPv6 and Wireless Networks

bemis sent us an article that talks about IPv6 and Wireless, and how the two seem to fit together pretty well. (Especially since at the rate we're going your home stereo is gonna need its own class C)

Comments

Need IPv6

[jabber01]

Private use may not necessitate v6, as long as we all have one IP address. But we don't.

Anyone with a cell-phone will need an IP address soon. Shortly thereafter, anyone with a pager. Then anyone with a car. Then anyone with a major appliance. It doesn't always make sense - is a networked washing machine really necessary? But as soon as there is someone willing to buy it, it will be made available.

Imagine, having a washing machine that can page you when it's done cleansing your tighty-whiteys? Now, unless you expect everyone in the world who wants a piece of this new technology, to set up their own wireless subnet, you'll have to agree that it's going to require a network which will support this sort of flexibility.

Personally, I'm all for a car that can self-diagnose, and wirelessly inform my mechanic/dealer of a problem. I'm all for email that will get routed to wherever I happen to currently be: work, home, cell, car..

The thing here is that v4 is running out of available addresses in a big hurry. A 32 bit address field just doesn't cut it anymore, and subnetting, masking, ghosting, shadowing, blah, blah and all other v4 hacks can only go so far.

v6 has a great deal to offer, but the acceptance curve is pretty steep. Not for technical reasons, but for financial ones. Networking hardware presents a significant investment, and well designed hardware happily continues to run when new technologies become available. It's very hard to justify the purchase of new hardware, if there is nothing 'wrong' with old hardware.

Incidentally, this is why very few offices are using fiberoptics, and so many are still on 10Mbps LANs. Stringing new wires (wires, nevermind routers, just wires) is a very expensive undertaking. Replacing old CAT-3 wires with CAT-5, to make 100Mb to the desktop possible (in an older office, for example) requires the office to be shut down, so workmen could gut the walls. Then you buy the switches and routers; and THEN you buy everyone a new NIC.

It's the COST, not the technical merit of the technology, that is keeping it out of our hands; and we are running out of v4 'tricks' very quickly.

The REAL jabber has the /. user id: 13196

IPv6 & Wireless

[jd]

I'm very much in favour of IPv6, which (through it's notion of transitional IP addresses) is (IMHO) much better at handling wireless scenarios, where the exact topology is indeterminate over any period of time.

That, I think, is one of the biggest road-blocks to stable, reliable, large-scale networks. The topology isn't going to be consistant, moment to moment, and IPv4 has no built-in mechanism for supporting that.

(You can use packet-forwarding, if all the nodes support it, but mangling packets and re-sending them gets messy and lossage is going to be higher.)

The day when it'll be possible to connect your portable via a wireless link when on the move and via a land-line otherwise, without EVER having to lose connectivity at ANY point, is coming.

Mobile, IPv6, & Real-World Latency

[maggard]

IPv6 is not going to take the world by storm.

Rather, it's gonna have to happen as a series of foggy-areas gradually coalescing into local showers.

Why? First there's complacency. All of the problems with IPv4 have been patchable, work-aroundable, or otherwise resolvable. There's no screaming need for IPv6 right now. There are theoretical benefits and additional features built in but no absolute pressing need for it today, tomorrow, next quarter or next year.

Second there's the additional cost. Developing, testing, deploying, and supporting IPv6 is gonna cost. Apple & Stanford both did massive IP renumberings a few years ago; they cost millions and that's much less difficult then switching IP stacks and network infrastructure. Anybody that rolls out IPv6 in a big way is gonna have to spend a LOT of money doing so and frankly I know of few budgets with that kind of slosh in them.

Software compatibility. Applications and utilities across the board are hard-coded to use IPv4. From word-processors to chat clients to multi-tier ERP applications they all expect IPv4 and burp & spit-up when fed IPv6. Yes there are work arounds and alternatives and all of that but it quickly turns into a rats-nest of slightly different applications and idiosyncratic configurations and the whole set-up just gums up.

Hardware support isn't there yet either. Few products support IPv6 yet. Fewer still do so well. Of those almost none do so optimally. From NICS to routers to management systems to contracts and manuals the boxes aren't ready yet. Sure for a lab or two, even a floor or two on a research building but the minute you start plugging in the obsolete, the unusual, the critical stuff you start running into problems.

That million-dollar super-printer downstairs? No go. The fancy networked building security system? Locks up solid. The black-box encryption system for routing email to our overseas branches? We don't know what happened but now all of the LEDs glow solid and we can't get it to reset.

IPv6 is deep and untested waters. IS/IT/MIS/etc. is complex enough these days without throwing in a giant wild variable like IPv6.

Furthermore we've been burnt before. Remember when OSI was going to rule the world? Then ATM was gonna take over. Now IPv6 is the heir apparent. Frankly until it's out there and in significant quantities that it's a standard order most folks aren't gonna touch it. Oh there will be the occasional test and we'll have a favored techie bone up on it as a cookie/insurance-policy but nobody is taking it seriously.

Even in the wireless phone world IPv6 is finding it hard to roll-out. The equipment is expensive, tolerances are tight, and the requirements are brutal. These are telephony folks - they still have the old tight-ass conservative Bell-ways trained into them even in the wild-'n-woolly new age of wireless. They want many-9's of reliability, flawless interoperability, and the ability to scale quickly and massively before they'll commit.

IPv6 looks great when you're hacking around on your home box. But when it comes to signing the check for a couple million dollars or more (mebbe much more) for hardware, support, training etc. and you know that this will have significant repercussions on your career it suddenly looks much less appealing. IS/IT/MIS/etc. executives aren't cowards, but to get where they are they have to be survivors. Right now & for the foreseeable future IPv6 doesn't look like a good bet to be making.

Finally - what are Cisco, Nortel, 3Com, etc. using internally? Ipv4? Uh huh. If it's so great why aren't the darn manufacturers "eating their own dogfood"? Perhaps even with all of the support in-house they know it's not worth running yet, even for the bragging rights.

-- Michael

ps On the other hand for students, developers, and the ilk - bone up, design-with-this-in-mind, this could be a sea-change that will make your fortune.