Eircom Netopia 2247 and IPv6

I've had a problem with IPv6 on my home network for a long time. With Vista\Win7\recent Ubuntu releases I found that ipv6 aware applications (like Firefox) would slow to a crawl and spend ages "resolving" addresses that I could use nslookup to resolve without an issue. Some digging pointed at IPv6 name resolution and I found that disabling ipv6 at the OS level or disabling ipv6 internally in Firefox using the network.dns.disableIPv6 setting in about:config fixed the symptom.

I've been playing with Chrome OS in a VM today and I found that it was spending an insane amount of time (apparently) resolving addresses so I started to look around and see if there was any way to disable IPv6 DNS lookups in Chrome\Chrome OS as that seemed to be a likely suspect. It appears that the developers are taking an ideologically purist approach to this and are deliberately not including a user switch to disable IPv6 because they feel that if they mask the problem at the user level then the incentive for providers to fix their buggy IPv6 infrastructure will be seriously diminished. That's a reasonable policy I suppose so I had to dig further into the IPv6 problems at my end.

The root of my problem turns out to be the Netopia 2247 DSL router that I got from Eircom. By default it sets itself up as a DNS proxy and it doesn't handle IPv6 AAAA requests well. The Netopia's web based GUI doesn't have any obvious way to disable this "feature" and all clients that get DHCP addresses from it end up with the router address as their single DNS server address. It does have Telnet management interface that gives you much finer control of the config though.

To fix the problem telnet into the Console - (download Putty if you're on a newer Windows OS that doesn't have a native Telnet client). Login using the same credentials you use for the web interface (the defaults are Username: Admin, Password is not set, just hit return). Then use the following commands:

configure

set dns proxy-enable off

save

restart

The router will reboot, this will force all clients to reconnect and their new DHCP settings will now point directly at whatever DNS servers it has been configured with rather than the router itself.

Chrome OS now boots and authenticates rapidly and there's no lag at all now while it tries to resolve addresses so I think I've finally found a proper fix for the problem.

You know you’re a sad Geek when..

Following some links about the new Gnome Network Manager included in Ubuntu 9.04 I found a blog by Dan Williams who was responsible for much if not most of the underlying work that has made it such a pleasure to use with 3G network devices.

This article where he complains about the firmware on one of the new Huawei 3G cards brought a smile. Been there dude! I often had exactly the same reaction when testing my own cellular device discovery code against various hardware. Then I realized just how sad it was to be able to fully understand and sympathize with a comment like:

…the response to AT+GCAP is simply “+CIS707-A, +MS, +ES, +DS, +FCLASS”.  No, it’s not prefixed with “+GCAP: ” like every other modem on the planet that I’m pretty sure the relevant standards (TIA/EIA/IS-131, TIA/EIA-602, and V.250) require..

I need to give this new Network Manager a proper try out on some diverse hardware now as it seems possible that it finally gets Ubuntu up to a level where it is as good an experience as that seen on the best Windows systems.

Everything you ever wanted to know about VMware vSwitches

I just came across a fantastic series of blog posts from Ken Cline at “Kens Virtual Reality Blog” that ties up pretty much everything you will ever need to know about configuring networking in ESX 3.5 \ ESXi. No doubt there will be some significant changes in ESX 4 but even so this is a great resource. There are 5 parts to this so far, all the earlier items which explain all the background detail are linked from this fifth part where Ken makes some hard and fast recommendations around how to tie it all together to create robust and effective ESX network design decisions.

The Great vSwitch Debate (Part 5).

He covers everything from the basics of creating vSwitches and Port Groups, VLAN tagging – clearly explaining what VGT\EGT\VST tagging variations actually mean and how they are enabled, load balancing options and their usefulness (or not) especially in relation to 802.3ad \ LACP\PaGP, NIC failover especially with regard to Beacon Probing and Link State Tracking on your physical switches, security options (Forged Transmits\MAC Address changes, Promiscuous mode) and a lot more. Definitely recommended for anyone who either builds or wants to understand ESX installations.

Blue Sky Engineering

Koenigsegg, those mad Swedes beloved of Top Gear make supercars that arguably demonstrate some of the most incredible real world engineering skills on the planet, seem to have lost the plot somewhat and started drinking some serious hallucinogens.

They’ve just announced an electrically powered “eco” supercar with the rather oddball name of “The Quant”. Possibly they’re making some in joke swipe at their usual customer base and this is just an early release of an April Fool’s joke but the numbers linked to this magical beastie are pure fantasy - “Blue Sky Engineering” as its called in the space business.

Some key "performance” claims and characteristics of the components they claim they have developed or that are near completion:

Top Speed 171mph, 0-62mph: 5.2 seconds , Range 312 miles, Power 381kW/512bhp / torque 715nm / 527 lb ft. Mass 1780kg.

So far so good. Those are probably doable in something shaped like a supercar although it’s definitely not getting 312miles doing 171mph speeds but they don’t specifically claim that so we’ll let them off.

The battery\energy storage system they claim they will use is called the Flow Accumulator Energy Storage (FAES) from some crowd called NLV. This amazing widget will provide the energy storage to facilitate the aforementioned 312mile range. Interestingly it also claims that it has a volumetric energy density of 600 Wh/l and can store a full charge in 15-20 minutes. The amount of energy it would take to hit ~300miles is going to be as near as makes no difference to the amount of productive energy produced by a petrol engine in a similar standard car with the same range. Let’s say that would be about 40 litres of petrol which has a volumetric energy density of about 9600Wh/l. Petrol engines are pretty inefficient so only about 25% of that potential actually gets converted to useful energy so that sort of range requires something like 90,000 Wh. To put 90,000Wh into a storage system in 20 minutes you have to push in 3 times that amount of Watts ie 270kW. That’s a lot of juice – even with industrial grade 380v 3-phase power you need to push almost 240 amps of current into it. Frankly I can’t see how you would be able build a power handling system into a car sized system that could handle that without using something like 5kV input voltage and that brings its own serious safety issues (not that 380V 3phase is exactly safe either but 5kV is distinctly dangerous especially when it would still be delivering 15-20A of current).  On the plus side if they actually have that miracle power system then they should be able to store that amount of energy in something like 200litres of battery which is nice – most current electrical cars currently need 5 times that volume.

Impossible? Probably not impossible but quite unlikely and certainly totally impractical today.

The part that sets my alarm bells going though are the “invisible..high efficiency Pyradian thin film solar cells” that (according to el Reg) has a conversion efficiency between 38 and 50%. I don’t think so folks. The very best hand made multi-junction monocrystaline PV cells barely hit the bottom end of that range and they cost a fortune by any measure. Thin film tech is currently languishing around with conversion efficiencies in the mid to high teens and I’ve never seen any evidence of serious work on multilayer (presumably multi-junction actually) thin film PV. Assuming they have this other wonder material though what good would it be? They say the whole car is covered with it, so at 4.8*2*1.3m that’s about 25m^2 of PV cells that could (in theory) yield about 1kW of power on a reasonably clear day in summer time at these latitudes. Even under ideal conditions (Sun directly overhead in California in Mid Summer) that would be worth about 3kW. That would make some sense – give it 8 hours in the California sun and you’d have enough juice to drive 100miles or so. Unfortunately, as I said, the PV conversion capabilities of thin film PV is less than half the claimed number and even in California you don’t get 8 hours of peak sun per day so the actual amount of power such a system would give you would probably only give a Quant owner a useful range of around 30-40 miles under ideal conditions and more like 5-10 miles per day in most places (on a good sunny day too).

I’m hopeful that I’ve got some of the numbers wrong here because I’d love it if this was all real but I suspect that it’s basically a better version of the Tesla, possibly with some technical improvements but without any of the tectonic technical advancements that are being claimed for them.

Better Management and Getting stuff Done.

I spent about four years as a manager at one stage in my career and it’s one of those things that I’m very glad I’ve done, as it was very rewarding in the end, but I’m even more glad that I don’t do now because it was probably the hardest, or at least the most stressful thing I ever had to do.

In my case (as I recall at least) – I generally scored average on internal managerial skills reviews. For a couple of years I tried to modify my behavior in response to my manager’s feedback in order to improve those reviews. That helped a bit (my manager was happier at any rate) but the improvements were small. I had pretty much accepted that I was an average manager and I’m not happy with average so I decided to go back to being an engineer, something I knew I could be great at. In the six months my transition back to being an engineer took I gave up on looking after my manager (and his manager and so on, hierarchy is a bitch) and decided to focus on working for my team. My final managerial skills reviews just before I left showed a dramatic improvement. That will be no surprise to any good managers out there, I’m sure.

Anyway the key thing in all this was that I learned a useful managerial trick that year – I spent my time looking after my team not (just) my boss, we all got a ton of really great stuff done, morale was way better and everyone (including my manager and the company) was better off. I also learned (finally) that you can actually learn to be a better manager, although I still think that really great managers need traits that can’t be taught.

This has all been a roundabout way to lead in to a great article I came across today that makes this very point and provides a very succinct set of guidelines on how to be a good manager. Truly great managers probably can’t be made from total muppet starting material but for the most part anyone having to “manage” people in anyway can become very good at it by following the advice here. For everyone else you should start demanding that your managers follow advice like this – life will be better for everyone if they did and a lot more stuff will get done.

Non Hierarchical Management – Aaron Swartz.

That reminded me of an earlier link I found. It’s basically an over the top call to create a culture of adventurous achievement by doing one simple thing – get stuff done rather than fucking about. It flies in the face of bureaucracy, “process”, safety and stability but bloody hell the world would be a much more interesting place if we adopted its basic idea – get stuff done. Wouldn’t it be great to work in a place where this was the “mission”?

Cult of Done Manifesto: – Boing Boing.

iPhone App Store Hits 500 million Downloads

Apple are now advertising that the App Store passed the 500million download mark recently which is a pretty impressive number no matter how you look at it.

I’m always suspicious of these sort of claims because (naturally)  any company will try to put as positive a spin on numbers like these as they can so I dug a little bit to see if they give a breakdown of what the number actually means – unfortunately Apple don’t really publish enough data to really figure out what that 500million actually means but there is just about enough other data out there to work it out for ourselves.

There are various estimates in the Blogoshphere of 16-18 million iPhones sold to date. That seems reasonable as there’s solid info from Apple that there were 4 million total to date at the start of 2008 and TMO showed that at least 10 million were sold or in the distribution channel by the end of September 2008 so 17 million seems like a good estimate of the total. There is very little info on the total number of iPod Touch sales but there are some hints – notably that the iPod Touch now outsells the iPod classic and that the average selling price for all iPods is in the $150-$160 range with a total volume of around 40million units for the year. A reasonable total number from that would be 6million or so iPod Touches for the year and something like 8 million in total. So overall we’re looking at about 25million potential iTunes App store users as of the end of December.

That would give us 20 downloads per user which is a remarkable number. Power users on high end Symbian\Windows Mobile phones would have downloaded some apps but on average a typical mobile phone user rarely downloads and installs any applications at all, for Apple to have created an ecosystem where the average user downloads 20 apps over a period of 6 months is incredible.

Looking at my own iPhone I have a total of 53 separate apps downloaded. Digging a bit deeper that breaks down into 18 paid apps and 35 free apps. Looking at the invoices from the iTunes Store it appears that I have a total of 74 downloads recorded which means that in addition the the basic downloads for new apps I’ve had 21 update downloads. That seems a bit low to me but it’s possible that it’s right and it’s definitely not significantly off the mark as most apps never seem to get updated. I’ve spent a total of €65.22 on the 18 paid apps. Now I’d say that I’m probably more likely than most to pay for apps so let’s assume that the average user (with 20 downloads) breaks down to 15 apps in total, 3 of which are paid apps with a total cost of ~€10 at an average cost of around $3-$4 each.

That should then mean that the App Store would have pulled in about €250m in 2008 with 30% ($75m) of that going to Apple. Those numbers are in the middle of the range of the $50-$100m estimate from Silicon Valley Insider at the start of December which was based on total downloads of 300m. I think they overestimated the percentage of paid apps (at 33% versus my guess of about 20%) but the average price is certainly right at $3 per paid app.

It’s useful to compare this reality to Eric Schonfield’s June 11 2008 TechCrunch article on projected iPhone App Sales for 2009 where he was tearing down a prediction of $1.2billion revenue from iTunes Apps in 2009 by Piper Jaffray’s Gene Munster. Schonfield correctly predicts that at least 70% of apps will be free and that the average price will trend towards $3 or less per paid app. He then also correctly estimates that the total number of iPhones by the end of the year will be around 16million and estimates the total iPhone sales for 2009 will be 25million at most. In fact the seems to think that the cumulative iPhone numbers by the end of 2009 will be somewhere between 25 million (ie 10 million new sales in 2009) and 50 million (the more optimistic but still realistic upper end estimates from other analysts) and implies that total revenues from the App Store are therefore likely to be between $150m and $300m. What he got dead wrong is that the average user appears to be downloading at a rate of around 30 apps per year and paying for around 6 of those. There’s no shame in that to be fair, I don’t think that any rational commenter at the time thought the App Store would be as popular as it has turned out to be.

I also think that he failed to factor in the iPod Touch which appears to be a very significant factor for the App Store and I personally I think the total number of iPhones\iPod Touches by the end of 2009 will almost certainly be close to 55m (the existing 25 million + about 30 million new sales in 2009). As we now have strong evidence that the average App sales number appears to be about $10 per device over 6 months (and it’s accelerating remember, not slowing down yet) so that total revenue projection of $1.2bn for the App Store for 2009 actually looks like a good bet right now, despite the recession. 

More On Windows 7

So after three days using Windows 7 in anger in a consumer sense I’ve got to say that it’s a fairly impressive offering for a Beta, it’s certainly going to become my default OS at home and probably at work too now that I have the option to switch to it.

I’ve been using it on my XPS M1330 at home in full Windows 7 mode, Aero interface with all the eye candy turned on and also at work in a VM where Aero doesn’t work. The difference in usability surprised me – I’ve always found Aero on Vista to be pointless overhead but in Windows 7 the eye-candy is actually useful and the UI improvements with Aero-7 make it a must have for me.

Speed wise and memory footprint wise it’s as fast as and maybe quicker than Vista on the same hardware (full boot+login to a usable desktop takes ~43 seconds from cold on my M1330 (2.2GHz Core Duo\4GB RAM), Vista takes ~44). Memory footprint for both Vista and Windows 7 on this hardware is about 1GB RAM with a pair of browsers (IE8 & Chrome) open. On the VM on my work laptop (a Dell Latitude D630 with a 2.2Ghz Core Duo and 2GB RAM) Windows 7 boots in about 60 seconds including logging in but uses only about 500Meg RAM of the 1GB assigned to the VM with the same two browsers running. Interestingly VMware Workstation 6.5’s “New VM” Wizard handles Windows 7 seamlessly giving a completely zero touch installation that took about 35 minutes to install from the DVD.

No crashes so far on either system and the only glitch has been that I can’t find a way to disable “Tap to Click” on the touchpad on the M1330.

There’s quite a nifty blog entry here  from Tim Sneath at Microsoft that lists a bunch of the new features which includes a couple of things I’d missed on Monday – most notably the CTRL+WIN+Left/Right arrow for moving maximized windows between monitors on a multi monitor display, ALT+CTRL+TAB for tabbing between windows within a single app, CTRL+Shift while clicking on an icon to launch with elevated privileges, WIN+Space for peaking at the desktop, Shift+Right click on a folder in Explorer adds Open-in-New-Process and Command-Prompt-Here.

Digging deeper I went scratching around for further changes within apps and came up with the following initial list for those curious about the command line environment that Windows 7 delivers. There are a few but on reflection the number of changes are quite small.

Robocopy:  Additional switch options not present in Windows Vista Version.

• /EFSRAW :: copy all encrypted files in EFS RAW mode.
• /DCOPY:T :: COPY Directory Timestamps.
• /SECFIX :: FIX file Security on all files, even skipped files.
• /TIMFIX :: FIX file Times on all files, even skipped files.
• /SL :: copy symbolic links versus the target.
• /MT[:n] :: Do multi-threaded copies with n threads (default 8).
•              n must be at least 1 and not greater than 128.
•              This option is incompatible with the /IPG and /EFSRAW options
•              Redirect output using /LOG option for better performance.
• /FFT :: assume FAT File Times (2-second granularity).
• /DST :: compensate for one-hour DST time differences.
• /XJD :: eXclude Junction points for Directories.
• /XJF :: eXclude Junction points for Files.
• /BYTES :: Print sizes as bytes.
• /UNICODE :: output status as UNICODE.

Netsh: Additional contexts not present in Vista version

• branchcache    - Changes to the `netsh branchcache' context.
• dnsclient      - Changes to the `netsh dnsclient' context.
• namespace      - Changes to the `netsh namespace' context.
• trace          - Changes to the `netsh trace' context.
• wcn            - Changes to the `netsh wcn' context.
• wfp            - Changes to the `netsh wfp' context.
• wwan           - Changes to the `netsh wwan' context.

Ipconfig: Additional ipv6 options, drops “compartments” concept

Diskpart: Support commands for Virtual Disks (Attach, Detach, Expand, Merge..)

Sc.exe: Adds support for service triggers \ trigger queries

Setspn.exe:  Register custom Service Principle Names in DNS

Tzutil.exe:  Enables scripted timezone changes

 

New Apps / Features:

Isoburn.exe: Compact GUI app for burning an iso to a CD\DVD

PSR.exe: Problem Steps Recorder.

Resource Monitor: Significantly enhanced version of the Resource Monitor that was previously available only from within TaskManager->Performance tab but is now also in the Start Menu.

Includes additional summary features for each heading using a tabbed interface in particular providing a very useful Physicial Memory utilization graphic.

Private Character Editor: Roll your own customized characters\fonts.

Admin tools now has a more logical home in the Start Menu -> Maintenance menu in addition to being buried inside the Control Panel.

Aero Shake: Pretty nifty clutter clearing – minimizes all other windows when you grab a window title bar and “shake” it. Additional gestures are supported elsewhere – left click+hold and then an upward “swoosh” gesture on the taskbar opens the right context menu (it’s more intuitive than it sounds and great on a touchscreen)

Smart Maximizing\restore\tiling by dragging a window to the top of the screen it automatically maximizes, by dragging it to the left or right edge it grows to fill a tile that takes up half the screen set flush against which ever side you drag it to. Dragging the title bar away from the edge restores the window to it’s former size.

Fonts Subsystem has been given a whole new look, no more windows 3.1 Add fonts dialog. Yaay.