Talk:IPv6/Archives/2004

Other features

This article seems to be entirely about IPv6 addressing. IPv6 has a bunch of other features, too, IIRC. Built in VPN support (IPsec?), QOS?, ...?

IPSEC is supported by IPv4 too, as is TOS. The main practical difference at the moment, other than address size, is better support for mobility. IPv6 does have a "flow ID" field, but no use it made of it at the moment. Noel 14:10, 5 Sep 2004 (UTC)

IPv6 "production" use

If you look at the current size of the IPv6 BGP routing tables here, you see that the size of the IPv6 routing table is between 150-600 entries, depending on which node you look at. Of about roughly 500 IPv6 AS's (a better measure of the topological complexity of the network), 3/4 of them only report a single route. Only 3 AS's are advertising a number of destination address ranges in 2 digits.

By contrast, if you look at the equivalent IPv4 data here, the equivalent numbers are between 140,000-180,000 routing table entries. The IPv4 system has about 18,000 AS's, and on average they each report 8 routes.

In other words, 10 years after the IETF anointed it as the successor to IPv4, and several years after MicroSloth made it availale as a standard part of Windows, the currently "operational" IPv6 network is roughly .4% of the size of the IPv4 network.

It's clearly not anything other than a large-scale testbed at the moment. When it gets up to, say, 10%, then you can say it's on its way to replacing IPv4. I chose the wording for the description of IPv6' status very carefully - at the moment it has been formally adopted, but its success is still an open question. Noel 03:11, 18 Sep 2004 (UTC)

I know about this data, but you are comparing apples to oranges. Let's start from the number of announced routes: on my network, where I have reasonably aggressive filters, I get from my upstreams 520 routes. The different numbers you report can be easily explained knowing that the 6Net view does not have 6bone routes and that the 6bone routing table is obviously a subset of the full view.

Comparing the number of prefixes announced by each AS is silly, because of the very aggressive aggregation goals driven by current RIR allocation and ISP filtering practices. The idea is that every AS should announce a single prefix, which will be large enough to aggregate all single-homed customers. Handling of multihoming is still an open issue, so you should also remember that the huge number of tiny AS announcing a single IPv4 prefix which exist in the IPv4 world has no equivalent.

Then, only about 60% of allocated production prefixes are being announced, so there is also a large number of ISPs planning adoption. Considering that half-decent IPv6 support has started to appear in IOS production trains only in the last year it's reasonable that adoption just started.

My ISP offers IPv6 access and has IPv6 transit customers, and many other ISPs and large carriers offer production or pre-production quality IPv6 connectivity. Because of this I do not think is correct depicting IPv6 networks as a "testbed". You are probably only aware of the US market, but in Europe and Asia IPv6 is more widely adopted.

Md 22:36, 18 Sep 2004 (UTC)

Look, no matter what the exact number of routes is (and I did say "between 150-600", which clearly includes the 520 you report), and even taking into account the hoped-for increase in address aggregation, the number is still clearly very small, compared to the IPv4 deployment.

As I also pointed out, the number of AS's is a better measure of the topological complexity of the network, and there are no features which would decrease them, and the number of those (~500 as of when I wrote) is again much smaller than the number for IPv4 (about 3%).

Future plans are worth the paper they are written on, and the same goes for predictions. I have a whole file of (incorrect) predictions by IPv6 proponents dating back almost a decade (although I have to say Jim Bound has been the source of the most, errr, formidable), and they all have the form of "well, we only got the <foo> last year, so now we expect things to really take off". So that excuse/prediction form (which you repeat) has worn somewhat thin.

Well, maybe IPv6 will be successful, and maybe it won't, but this is an encylopedia that reports facts, not the results of crystal balls. or marketing hype. IPv6 deployment today, over 10 years after it was selected, remains very small, and those are facts. If you don't like my phrasing of "its proponents hope it will form", I will simply insert the facts instead, and let them speak for themselves. Noel 13:27, 29 Oct 2004 (UTC)

PS: It's precisely because IPv6 still doesn't have a standard multi-homing mechanism (although the WG is making progress - it has recognized that the only archictually realistic solution within the current routing architecture is to use multiple addresses) that any people who currently want to do multi-homing with IPv6 have no choice but to do it the exact same way as it is done in IPv4 - which is to globally advertise their own local prefix. If that is not happening, then what it means is that those people using IPv6 are not doing multi-homing, which is simply further evidence that it's not being using in a production environment, which needs robust connectivity. Noel 13:30, 29 Oct 2004 (UTC)

Personally, I'd have to agree there is insufficient evidence to prove IPv6 is gonna succeed. Having said that, IMHO, it can't fail. China, India and so many other countries which are scourging for IPs are going to ensure that. As we see, China, Japan, SK are already making plans and they have the ability to implement them. Maybe Europe and NA won't follow immedi but then as Asia, then SA and Africa begin to dominate, they won't matter so much.