HP proprietary versus commodity hosts: Unix v. MPE
Web console resets, environment rebuilds, dumping form printers lead Hidden Value

IPv6: no MPE/iX support, and no matter, yet

One month ago this week, the latest generation of Internet protocols celebrated a kickoff week. Another one. IPv6 never made it onto the MPE/iX playing field. But some of the sharpest network gurus for HP 3000s say that the new Internet Engineering Task Force (IETF) standard isn't being baked into enterprise network name management. The fact that there's no IPv6 doesn't matter to 3000 managers -- at least not yet.

760px-Ipv6_address_leading_zerosIt's been a long time since IPv6 started its march to relevance. The Internet was built from its roots in the 1980s on the 4.3 billion IP addresses of IPv4, but the IETF estimated that 2011 would be the year when every address would be used up. So this next-generation IP standard was first approved about the time HP was releasing MPE/iX 5.0, in 1998. It's reasonable to think 14 years would be enough time for the world's computing community to embrace a crucial extension of IP addresses. But in the real world, IP addressing is a lot like HP 3000 deployments: What's in place isn't broken for many people, so there's no clamor to replace it.

Jeff Kell is a wizard of networking for the systems at the University of Tennessee at Chattanooga that host the 3000-L mailing list. He notes that "the 3000 of course will never do IPv6, short of some major overhaul patch -- and due to the extent of it, that seems impossible." But he adds that enterprise servers, from HP 3000s out to the ubiquitous Linux PCs, aren't making a move to IPv6 in any significant number.

Other than some of the much-publicized "World IPv6 Day" experiments, most of the "server" side of the network remains IPv4. To account for the imbalance, most IPv6 client areas have some sort of gateway, emulation, or tunneling solution to maintain connectivity with IPv4 services.

Kell goes on to explain that the IPv6 installations so far "are primarily focused on the 'client' side of the service equation, where the uptick in connected devices is growing at the greatest rate, and where IPv4 address depletion has actually become an issue."

The issue that emptied the world's bucket of IP addresses is similar to the one that is holding the 3000's clock at a 2028 reset point. Internet addresses, made up of four sets of three-digit numbers, are clamped inside a 32-bit design. Vint Cerf, one of the fathers of the Internet, said that IPv4 wasn't really intended to be a working implementation of addressing. But after engineers argued for years, Cerf said it was time to make US tax dollars spent by the Defense Department on ARPA pay off.

After a year of fighting, I said—I'm now at ARPA, I'm running the program, I'm paying for this stuff, I'm using American tax dollars, and I wanted some progress because we didn't know if this was going to work. So I said: OK, it's 32-bits. That's enough for an experiment; it's 4.3 billion terminations.

HP 3000 customers, and anybody else who lived through Y2K, know how this kind of plan works out. Two digits were supposed to be enough to describe years. The 3000 CALENDAR intrinsic, written for the most senior MPE Segmented Library (SL), uses only 7 bits to describe which year is in effect. That delivers a maximum number of 127 years which you can express, and MPE was built with 1900 as its base for dates. 2028 is the first year when date stamping becomes inaccurate.

The situation in the IPv6 world has a 3000 echo to it. HP 3000s won't stop running in 2028, even though their calendars will reset. Dates will be off, but it's not a System Abort situation at all. In the same way, IPv4 is likely to have gateway resolution help for many years to come, so those 4.3 billion terminations continue to work.

IPv6 is still fighting for some relevance among large Internet service providers, Kell says.

It is still quite difficult to find end-to-end IPv6 examples in the real world without some intermediary workaround. There are still no backbone providers that can offer you native IPv6 connectivity to all of the current IPv6 networks (it still requires at least two, last I checked).  Very few home / SOHO switches/routers have native IPv6 available, and hardware support is an even smaller subset.

Kell also pointed out that default IPv6 support on client operating systems like Windows has ugly artifacts, so far. Transition technology that runs off something called a Teredo public gateway becomes a time-wasting stop for Internet traffic on Vista-and-later PCs.

So what is driving IPv6? Windows (since Vista) and MacOS (10.5+) both default to making desperate attempts to find some minimal IPv6 connectivity via a number of gateway and tunneling protocols. One of the most inefficient imaginable is Windows and its Teredo tunneling. If Windows cannot find IPv6 connectivity, it will try to connect to one of Microsoft's public Teredo gatways and bounce all  of your subsequent network traffic through there.

It is inefficiencies such as this -- combined with the fact that each of the operating systems will prefer IPv6 over IPv4 if they find any connectivity -- that may haunt system managers of all sorts (not just 3000 owners) as you try to debug and troubleshoot client connectivity issues.