If you’re wondering what comes next for US space flight, I can suggest reading the articles entitled “WHY THE US CAN BEAT CHINA: THE FACTS ABOUT SPACEX COSTS” and “TAKING THE NEXT STEP | COMMERCIAL CREW DEVELOPMENT ROUND 2″ on the SpaceX website.

Actually that request, as well as the one that I thought was from MWEB, were both from hosts on their networks connecting via tunnels (6to4 for WA, and Teredo for MWEB). Even worse, I suspect both tunnels terminated on overseas gateways.

So, for the second time this week, I’m eating my IPv6 words. MWEB and Web Africa, I un-congratulate you.

Because both tunnel mechanisms embed the source IPv4 address in a IPv6 address, whois(1) is able to extract the real endpoint address and do a look-up on that. It actually reports that it has done this conversion, but I wasn’t reading the output carefully enough. I’ll try making up for the error with interesting (and hopefully correct) commenatry about the state of IPv6 in our nation.

Although Localloop didn’t see evidence on Wednesday of native IPv6 turn-up in South Africa, I believe there is hope for this to change in the near future:

About 54 South African networks have IPv6 prefixes allocated to them, but only 24 of those prefixes are (currently) correctly routed on the IPv6 internet. Many of them have never been routed on the Internet. For more detail check out this excellent resource:

http://www.sixxs.net/tools/grh/dfp/all/?country=za

Among those that are working, you’ll find big players like Telkom SA and Internet Solutions, both of which I’m sure would sell you expensive “corporate” IPv6-enabled access over leased lines. Most interestingly however:

• Web Africa’s block has never been announced according to GRH, so I was very wrong about them rolling out IPv6.
• MWEB, on the other hand, is announcing an enormous block, a /26. This is twice as big as IS’ block, and 64 times (6 bits) bigger than most other blocks (Telkom’s or TENET’s for example), yet their users are not hitting my site from this prefix.

To be fair, local ADSL ISPs like MWEB and Web Africa have a good excuse for not yet providing native IPv6 to their users: as far as I know, it is impossible. The cause of this problem has to do with the routing set up (“IP Connect”) that Telkom SA uses to re-sell access to their ADSL cloud.

The good news is that Telkom has a new system in the pipeline called “IP Stream” that would work around this limitation by providing bit-stream connections between ADSL users and their ISPs. I believe some providers are getting ready to take advantage of IP Stream to roll out IPv6 to users en masse. If I’m right, it would explain how MWEB succeeded in motivating for a /26 to AfriNIC.

With some luck, next year’s World IPv6 Day will see Localloop getting some (real) native IPv6 hits from local consumer ISPs.

It was, however, a successful experiment. A few things I learned:

1. I actually got some IPv6 visitors! I was not expecting any, never mind South Africans coming from ISP-routed blocks, so this is great to seeI got some IPv6 visitors:
   • One 6to4 tunnel visitor from MWEB
   • One Teredo tunnel visitor from Web Africa
   • One HE.net tunnel visitor (Hi Aragon!)
   • A number of native visits from IPv6 Google crawlers
2. Despite the TTL on my A records being 1 hour, I continued seeing hits over IPv4 the whole day, which seems to indicate that some systems (especially web crawlers) seem to ignore the TTL when caching DNS records.
3. Running single-stack IPv6 is non-trivial and involves more than just your web server.
4. We can conclude that having a world-day with the objective of running dual-stack is not that useful.

So thanks for participating. It was fun. Kudos especially to MWEB and Web Africa.

Next year I’ll do things differently:

1. DNS will definitely be fixed, and I’ll also be accepting email over IPv6.
2. I will go single-stack IPv6 on my client side also (for the day)
3. In addition to removing A records, I will also block IPv4 packets (on the web server)
4. I will investigate the feasibility of disabling IPv4 in the kernel
5. Maybe I’ll also run DNS and email single-stack :-)

Update: The original post contained some errors, explained here.

Tomorrow is World IPv6 Day.

The big deal (apparently) is that some providers will be temporarily running IPv6 in parallel to their IPv4 services.

The protocol has been around for 15 years, and now they’re have a whole “world day” to turn it on, and then they’ll turn it off again.

Cowards.

For tomorrow, I will be turning IPv4 off on the server hosting this blog. Thanks to Neology, Localloop be reachable over native IPv6.

Some of their hardware is pretty good (I’m writing this on a splendid Dell Latitude, which I would choose over a Mac Book any day). Other hardware, not so good: the LSI RAID controllers (megaraid) Dell use in their PowerEdge server line – they’re just rubbish.

Sometimes I complain about these Dell annoyances, and wax lyrical about the smooth experience I had with HP Prolaint servers at my previous job.

Then, last month I loaned an HP Proliant DL360 G7 from a colleague to use as a test system. Good hardware, but not significantly better or worse than the equivalent Dell box.

Standard on both is integrated out of band management, which provides the ability to interact with the server remotely (control power, access the console, connect virtual media). Dell’s is the iDRAC, and HP has the iLO. Since the feature is not optional, you always to pay for the management hardware.

It turns out that HP’s iLO requires an advanced software license to enable the remote console/virtual media feature. What I find surprising is that this is the most basic feature of out-of-band server management. Without it, iLO is of no use to me. Either my previous employer always paid extra for the advanced license, or they introduced this idiocy since I last used HP servers.

Anyway, I wanted to test some software, not fiddle with servers, so after learning from Roelf that you can download a 30-day evaluation license, I forged ahead, enabled the advanced features and got the job done. The next day I decided to re-install the OS on the server, only to find that the iLO decided that my 30-day license has expired.

So, I log a support request with HP using a web form on the same page where I requested the license:

From: Hewlett-Packard@g1u1081c.austin.hp.com [mailto:Hewlett-Packard@g1u1081c.austin.hp.com]
Sent: sábado, 16 de abril de 2011 05:40 a.m.
To: pdl-pdapi-L1@hp.com
Subject: [US] download questions – contact us

1. Describe the issue (please provide full details).

I requested and downloaded the 30 day HP iLO Advanced evaluation license (HPICEVAL), and installed it last night.

This morning the iLO reports: Evaluation license is expired, and advanced features are unavailable. Entering the key again gives the error: Could not install license. Demo license previously installed..

Soon I get a response:

Hello Simeon

Please refer to the following link for assistance with licensing questions.

http://licensing.hp.com/slm/welcome.slm

Thank you,

Kenryu Milko Yanome.

Ok, fine, I follow the link and submit the same request, and the next day I get another response:

Hello Simeon,

Please note that requests concerning Insight software are not supported by codeword_europe@hp.com mailbox.
I would like to ask you to resend your request to the following e-mail address: entitlement.support@hp.com. My colleagues will be happy to support you.

Best wishes,

Tomasz Huber
HP Licensing Support
HP Licensing Entitlement

Getting a little impatient now, I send the same request to entitlement.support@hp.com and receive no response.

I wonder, if a were to pay for the iLO Advanced license (so that I don’t have to lug a screen and keyboard to the other side of our campus, and sit in the cold server room while I wait for the OS to install), what are the odds that it won’t malfunction again and decide my license is invalid?

And if that were to happen, would the support provided by HP be any better?

Dell’s iDRAC just works, and at no extra cost. Why must HP punish me for trying to use their hardware?

Now here is the thing about Skype, it always works, unlike SIP and friends.

You see, on the modern Internet, there are so many middle-boxes interfering with traffic between users that only two kinds of protocols survive: those that have been in use by everyone since the Internet got commercialized (SMTP, HTTP) and hence no-one dares to break, and those designed by hackers trying to subvert the system (Peer-to-peer file sharing). Skype happens to be one of the latter (Skype’s P2P design was based on Kazaa).

This explains why my multiple attempts to get simple VoIP with SIP signaling working reliably on networks where I control the middle-boxes, all failed. Apparently even people who build particle accelerators can’t build something that works as well as Skype (it turns out they’ve tried).

So, I don’t expect Ekiga or Google Talk to work as substitutes any time soon. I would welcome an open alternative as robust as Skype, but I also don’t believe this can be achieved by following standards, it would have to be hackish.

Welcome to the Internet, it is an ugly place.

Now, the overlords in Redmond paying Nokia to get Windows exclusively onto a major mobile platform and soon after buying Skype is no coincidence.

I think they want a messaging service that can compete with RIM’s BBM, and buying Skype simultaneously gets them some technology to start with, as well as a large (170M) user base. The user adoption problem can be hard to crack when you’re slow-moving and lack innovation.

Unlike Roelf, I believe MS is likely to continue multi-platform Skype client support, but I have to point out that it wasn’t that great to start with. In my experience the Linux client is awful and I understand that it lags substantially behind the Windows client. I think this will also suit them: by providing some support for a wide range of platforms they benefit from extra network effect, but if you want the features, switch to an MS platform… resistance is futile.

“It is Telkom’s contention that the wireless local loop these days is more than a substitute for both voice and broadband communications,”
- Techcentral’s Candice Jones quoting Telkom’s chief of corporate governance, Ouma Rasethaba.

I doubt they believe that (which wireless local loop technology available today is as scalable and robust as ADSL2+ for the provision of fixed broadband?).

In fact, if I were Telkom, I would downplay the value of the copper local loop and encourage my competitors (like these guys) to invest in WLL alone for home/SME customers instead of a mix of mobile and fixed line infrastructure.

Following a detailed evaluation against stringent technical and operational criteria, ITU has determined that “LTE-Advanced” and “WirelessMAN-Advanced” should be accorded the official designation of IMT-Advanced. As the most advanced technologies currently defined for global wireless mobile broadband communications, IMT-Advanced is considered as “4G”, although it is recognized that this term, while undefined, may also be applied to the forerunners of these technologies, LTE and WiMax, and to other evolved 3G technologies providing a substantial level of improvement in performance and capabilities with respect to the initial third generation systems now deployed. The detailed specifications of the IMT-Advanced technologies will be provided in a new ITU-R Recommendation expected in early 2012.

– ITU WRS-10 press release

Two comments:

1. MWEB already peers with MTN. Here are the two hops between their networks¹:

   6 g-0-3-vic-jinx-2.mweb.co.za (196.22.163.1)
7 mtnns-2.jinx.net.za (198.32.142.31)

   I think what MWEB probably means by "transit" is that MTN charges them for peering. If MTN agreed to MWEB's ultimatum, they would still be peering at JINX. So if you are an MWEB customer who likes to surf Mybroadband.co.za², you should demand a discount to compensate for the de-peering.

2. "We do not believe in paying for transit or selling transit. Rather, we have invited all of them to peer with us."
   -- MyBB quoting MWEB ISP CEO Derek Hershaw

   What makes buying transit in London different from Johannesburg?

1. traceroute from MWEB uncapped ADSL in Pretoria – thanks Ed
2. for example - MyBB is hosted at Hetzner, on the MTN network

Providing uncapped Internet is like inviting 10 buddies to a braai (barbeque) at your house. You decide to cater for one steak per guest, but one of them tries to eat 8 of the 10 steaks. So now you have 3 options:

1. Ask the big eater to leave
2. Let him stay, and allow the other guests to suffer
3. Limit each guest to one steak

The implication here is that the uncapped broadband user who tries to flat-line his service all month long is like the greedy guest, and clearly the only way to maintain harmony is option 3 (the data volume cap).

He may have had some of the audience convinced, but this analogy is inaccurate. Here are some facts:

• The steak is not free, the guests are customers, and they want what they pay for.
• You’re advertising your service as uncapped, so the amount of steak is infinite.
• You’re advertising your service as broadband, so the steak is available at a finite, but high rate (say, one steak per hour).
• Your service is always-on, so it should be available all the time, at one steak per hour.

So, is the big eater being greedy?

If he is the only user, there are no fairness concerns. If he is sharing the service with 9 other users, fair queuing can be used to ensure each guest eats no more than one steak per hour.

So what is the problem?

In reality, consumer ISPs assume that users will take turns eating steak. This assumption holds true for the majority of users, so the ISP only caters for one steak per hour for every 10 customers. This is called over-subscription. For simplicity’s sake, let’s assume the turn-taking is perfect (turns of equal length, one user at a time).

Now the big eater that doesn’t wait his turn breaks the over-subscription model, because if someone else gets hungry while he continuously consumes 1 steak/hour, both users get 0.5 steak/hour, and both complain about slow access. Also, with flat-rate billing, the ISP might be making less profit (or even a loss) when serving big eaters.

In a previous post, I described how our industry came from the dial-up era where this turn-taking behavior of users was enforced by the high cost of telephone calls. The introduction of broadband increased the demand, while their input costs remained the same, so to provide a seemingly-broadband experience at a price the consumer could afford, ISPs introduced caps to enforce turn-taking.

Now that some of the input costs have decreased, uncapped broadband has become both affordable and of reasonable quality (given the correct amount of over-subscription), but uncapped service also removes the turn-taking restriction, so some ISPs who haven’t yet figured out how to adapt are unable to cope with “abusers”.

Of course, “abusers” are just users who want to use what they paid for. This makes the ISP who doesn’t provide what they promise the greedy party.

How can they fix this?

“We teach people how to treat us.”
– Phillip C. McGraw¹

A start might be to be honest in marketing and transparent about over-subscription. Are you encouraging your users to flat-line their links by calling your product “all-you-can-eat broadband”, and then complaining when they do?

Decide what you want and then communicate this to your users in clear and simple terms². For example, if you want your customers to stick to a specific usage pattern on their uncapped account, specify the pattern (eg. duty cycle of x%) in the product specification, and then enforce that pattern. Finally, don’t enforce a usage pattern if you haven’t specified it clearly, because your users will perceive it as bad quality.

1. I honestly never thought I would quote Dr Phil on my blog
2. Having a vague fair use policy does not count