Please be aware! This e-mail is liberally mixed with my personal opinion as well as technical fact, so read with that in mind. If you will be annoyed or offended by a little (or perhaps a lot) of ranting I suggest you skip this mail. Want my opinion of the modern ISP, read on...
> I did a quick Google in an attempt to find the benefits that IPv6 has > over IPv4. There does not seem to be a killer app for IPv6. The thing > that impressed me is the ability for a IPv6 node to be mobile and maybe > with the prevalence of wireless ISPs IPv6 will become a reality. I also > read that IPv6 is more efficient to route, so what does that mean in > percentage terms? Do you think it will be a good idea to start IPv6 > networks from the ground up, instead of starting with IPv4 and > upgrading to IPv6? That's the biggest sticking point... For IPv6 every app is the killer app, and that's actually harder to sell than you may first think. Secondly, there's no regulator. That's great for decentralisation and equality (sort of, and even so only in some cases) but it means there is no-one to mandate an update. And as most executives are cowards (if you think you're worried about your job, think of how much of a dork/tool/knob your CEO is and keep in mind that he probably knows he is out of his depth, and worries constantly that he'll be found out) so it's not like some turkey like Trujillo is going to risk his job over a new protocol, especially when no-one else is doing it. That's 90% of a CEO's job these days, it seems: don't do anything risky, and only do what has succeeded before. And if you're not a technology company, well then it's likely that the CIO is some accounting recycle or, worse, some boob from Marketing who wanted better oversight of the website so he sleazed his way into the role. I even saw a HR migrant once. HR. It's a nightmare only previously hinted to in the panels of Dilbert cartoons. Everyone has gotten so used to bandaids and hacks that we've missed the true benefit of IPv6: rationalisation. We were running out off addressing, IPv6 was taking too long and so we got NAT. NAT sucks. Sure, there are some "security by obscurity" benefits, but the internet was a peer network to start with. I've been fortunate enough to participate in a community network in my area, and I have come to love having a range of static addresses I control. And part of making a not-for-profit network work was IPv6, because it gave us addressing to burn. The infuriating issue we have struck, specifically, is acquiring internet routable addressing, because APNIC and the like cling to the old ideals of a tiered internet and limited resources, when, for the 'net as a whole to take the next step, resources have to become a secondary concern through them being abundant. But I digress... IPv6 turns packet level encryption from a PITA hack to core feature. QoS is a design priority and not an afterthought. RSVP is not fun. IPv6 is engineered, it works and has (almost) stopped changing under our feet (wasn't the evaporation of site local addressing a pain?). There may be something to be said for a hypothesis that suggests that, for some carriers at least, IPv6 will be the writing on the wall when it comes to the easy days of making a buck on the Internet. All of those issues aside, it's hard for a small time developer, like open source programmers tend to be, to set up a simulation network to really do appropriate dev and testing. That's the real trick. No one is writing code that does more with IPv6 than it does with v4. It's not like anyone is writing software that does a/b/c with v4, but d/e/f works fully under v6, and only in a limited fashion in v4. People just don't see the point of writing features that only 2% of geeks use, let alone the general population. And developers worry that if they tell customers they have to deploy IPv6 to run their software at its full feature set then customers/users might look elsewhere. As a developer, you're better off taking the easy road: make sure it opens a :: socket and can handle AAAA responses from DNS, and nothing more. There is no use coding in that cool QoS variable rate stuff, because no-one wants to use it and you can't test it. > > It only will really become meaningful when it becomes the default > > option from the major ISPs and carriers for the carriage of IP > > traffic. And even then I would suspect that for the most part, > > end-users will be able to choose to be shielded from the intricacies > > of IPv6. > > End-users for the most part use domain names so this wouldn't change. But those nasty BIND4 servers out there need to go... And any NT4 DNS servers, but if you're still running NT you deserve what you get. Anyone know if Netware 4 supports IPv6? > > Even in the briefest look around, you will realise how far away > > real-world adoption really is. Most configuration dialogues and web > > forms today still try to parse/display an "IP address" as a dotted > > quad (eg 1.2.3.4). That is, they ignore the IPv6 format of IP > > addresses. The other simple matter is that I would wager that if you > > took 10 experienced (10+ years) network engineers and ask them about > > IPv6 and what it is about and have they even played with it, 9 of > > them would probably have told you about the huge new size of the > > address space, but the need for it has been pretty well removed with > > private IP address space (10.x.x.x etc) and NAT, and no they haven't > > played with it. IMHO it really still is only of real interest to > > propeller heads (myself included). True. I myself just coded a bunch of address filters in milter-regex and none of them were IPv6 kosher, so everyone's guilty of it, I guess. > From what I read, IPv6 has been rethought and re-engineered so that it > just works a lot better. The real benefits of IPv6 cannot really become useful until it is in use, and everyone is currently sitting around waiting for the 'killer app', as you so eloquently put it. Use will create demand, and to do it any other way seems to be resulting in a very drawn out process. > > On the other hand I do believe it will come into play at some stage > > (because as Vint Cerf once said - every light buld will need an IP > > address in the future) but my current hunch is that it might be at > > least 5 years before any credible moves need to be made (by us end > > users). > > Maybe ubiquitous wireless mesh networks might do the trick... Community networks are still around, but current tech is not too good for Wireless MANs, though if WiMAX gear gets cheap that could change. We'll have to see how the market reacts to it. IPv6 is actually essential to the running of the community network, as it is the primary transport and the IPv4 traffic travels in IPv6 tunnels. We use multicasting for task like cache peering, video streaming to ntp time syncing, and we use the traffic management to keep essential network traffic like OSPFv3, BGP and SSH packets flowing even if there is high demand. Sure, you can do all of this with IPv4, but the addressing available in v6 lets our dreams run wild. And if you have every tried to bridge 802.3 QoS to RSVP you would want another option... I am not saying RSVP doesn't work, but that doesn't mean it's easy! > > I'm happy to be contradicted. > > I'm not really contradicting, just wanted some peoples thoughts on the > matter, because I have been ignorant. Vendors are a real culprit, but not the Cisco's (as much as I would personally like to lay the woes of the world at their feet), Nortel's or Extreme's of the world. It's the Linksys's (blah, blah, Cisco), Netgear's and D-Link's. Where are the IPv6 settings in the print servers, webcams and other devices that are common in networks? Why no IPv6 on Kyocera or HP printers? How much effort does it take to code? They don't even have to do the work themselves, frequently. The Linksys WRT54G and GS are a great example. From the factory they have no IPv6. But load on OpenWRT and you have every network option of a full Linux host, within the limitations of 32 MB of flash and the same of RAM... and a 230 MHz Broadcom CPU. Even so, people have created [EMAIL PROTECTED] setups on the WRTs, so they are proving to be an impressive unit. We will have to see whether Cisco lets Linksys keep making them. I think that maybe all we can do is lead by example. Build our own networks and give developers a place to work. Create demand and advocate. Everyone, pick a random day in the next two weeks and e-mail you ISP's support desk about IPv6. Mention you see that their DSLAMs are already capable of supporting it: it says so in you ppp logs for pppoe. Go ahead, check it out for yourself. If you have IPv6 setup and compiled into ppp and RP/whatever pppoe then you'll see both ends try a little IPv6 action. If you live in a high-rise building and you know other geeks in the building, here's a naughty but tricky way to set up a network in the building. This idea was thought up by another guy who used to be involved with the network here. Everyone get an access point of the same brand that does ad-hoc client, but cheap. Get a single ended pigtail for the AP from your favourite wireless supplier (or make one yourself, I'm about to tell you to solder something anyway). Solder (see) a standard TV antenna jack onto the open end of the pigtail (just the tip, not the shield!) and configure the AP. Plug into wall, wash and repeat for neighbours. Who needs EoP?!? Certainly fast enough for leeching and gaming. And if everyone has an ADSL link you can all setup squids and ICP them together. Or, come up with some exotic load balancing scheme to aggregate the DSL links together and I am sure that between IPv6 and you you'll get something going. Anyway, this has been a rambling pile of, well, stuff. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
