When there's only one

...there's only one choice
Everything here is my opinion. I do not speak for your employer.
June 2013
August 2013

2013-07-01 »

Remember, folks, every time you use a mock, you reduce code coverage. So you'd better have a good reason.

2013-07-02 »

"> there could well be numbers of old, 32-bit Unix
> system running stuff when the 32-bit clock
> counter rolls over and we find ourselves
> back in the 1970's.

Oooh - you mean when we went to the moon,
and you could fly supersonic to New York!"

(seen on G+)

2013-07-03 »

The level of standardization of Linux wireless drivers is, if possible, even worse than the level of standardization of Linux audio and printer drivers.  Actually, Linux video drivers are also pretty much unstandardized.  Wait a minute, now I've forgotten which drivers I thought had any kind of consistency.  Serial ports?

2013-07-04 »

systemd: worst idea ever, or merely the worst idea I can remember?

2013-07-05 »

Here's a (somewhat vendor-biased) pretty good introduction to wifi beamforming.

http://www.cisco.com/en/US/prod/collateral/wireless/ps5678/ps11983/white_paper_c11-722622.html

In short, Cisco claims 7-12 dB of signal strength improvement (in both upstream and downstream) through their beamforming stuff, and the 802.11n client only needs one antenna and no explicit beamforming support.  Which is a pretty big deal, if I'm interpreting correctly (10 dB should be equivalent to a 10x improvement in signal-to-noise ratio). I'm not really clear how this compares to any other/non-Cisco wireless modules, or how common beamforming is, or what.  One of the points of this paper is that some multi-antenna setups don't make very good use at all of the multiple antennas, compared to this DSP-based beamforming stuff.

That said, this DSP-based beamforming concept translates into a really big sensitivity to software problems.  It's not very hard to wire up some antennas, but a tiny little error in the math formula could completely destroy your beamforming performance, which means a 10x reduction (if your beamforming is ineffective) or worse (if your beamforming weakens the signal by accident) in SNR vs. the expected performance.  And testing this stuff must be absolutely crazy.

This may have everything or nothing to do with our GFiber wifi signal strength issues.

I think I'm in desperate need of some kind of remedial wifi signal strength training.  Does anyone around here know about this stuff?

2013-07-10 »

That sinking feeling you get when you realize the annoying bug you've been working on isn't the cause of your problem.  Or rather, it is, but so are a dozen other annoying bugs, not all of which you know yet, and some of which involve the manufacturing process, and any one of them would be sufficient to cause your problem.

In other words, that sinking feeling you get when you re-discover, for the nth time, that Engineering Is Hard.

2013-07-12 »

This article is very sad to me:

http://www.heraldextra.com/news/opinion/editorial/lessons-learned-from-iprovo-and-google-fiber/article_80f1393a-278f-59ce-88a3-7533d5ceb2ca.html

Because they are a combination of right and wrong.  Background: my home town of Thunder Bay, Ontario, Canada, has a city-owned Internet/TV/Phone company called Thunder Bay Telephone.  This is a rarity in Canada, so it's easy to compare them to other cities, and the difference is striking.  Thunder Bay Telephone is independently profitable (ie. by existing they reduce your city property tax), and they charge lower rates for Internet, TV, and Phone than the national alternatives, who are basically shut out of the market because their prices are higher.  We know their prices are higher because they are higher in every other city in Ontario, where they are essentially the only option.  So that's a resounding example of a noname city taking responsibility for their communication infrastructure the same way they take responsibility for their power, electrical, road, and garbage collection infrastructure.  It worked.

But anecdotes are not statistics.  There are also lots of government attempts to run Internet stuff that have failed financially, and iProvo is apparently one of them.

The underlying truths of the matter is that everything requiring a wire/cable/connector/road into each house in a city is expensive, and inherently monopolistic because once someone has run a wire, the cost of competing with them - running a second and totally redundant wire into the same house - is punishingly high.  In my opinion as a hippie-commie Canadian, inherent monopolistic structures like this ought to be regulated rather than just handing it over to a monopoly.

The flip side to the "we should have known better" in this iProvo article is that everyone else in America should have known better than to hand over a monopoly/duopoly on their Internet service to one or two private companies in each city.  What happens?  Slow/crappy Internet at high prices with terrible customer service.  Shocking, right?  But at least you're not wasting your tax dollars on it.  Although it's hard to see how wasting your tax dollars is much worse than just paying exorbitant monthly fees.

Ironically of course, I'm working on GFiber, which is another private company trying to do that thing I just set is punishingly expensive, that is, running redundant new Internet wires into everyone's house.  If we succeed where iProvo (and all of Canada) failed, then I guess I was wrong, and private industry beats government on this one, at least for now.

So I guess that means I get to either be right or be successful.  When in doubt I tend to aim for the latter. :)

2013-07-13 »

Preliminary wi-fi access point testing results: $30 Edimax wi-fi repeater has surprisingly excellent performance yet is surprisingly non-fiddly.

http://www.amazon.com/dp/B00794RDVY/ref=pe_385040_30332200_pe_309540_26725410_item

Ubiquiti devices have unsurprisingly excellent performance and are unsurprisingly fiddly.

2013-07-14 »

More preliminary results:

Fancy beamforming stuff and 1000 mW maximum output power in Apple's Airport Extreme 802.11ac does not seem to measurably improve maximum range vs. their Airport Express 802.11n with lower maximum power.  That makes me sad.

On the other hand, inside that range, something in there is clearly improving my connection speed.  I get a full 300 Mbit/sec for a fair distance from the device, and a generally higher transmit rate across the whole functioning region.  Since my Macbook doesn't even support 802.11ac, my guess is this is some combination of better antennas (there must be some reason that thing is so gigantic), higher transmit power, and beamforming features.  (In retrospect I guess it makes sense to me that beamforming can't help much at long distances. But it saddens me anyway.)

So far both Airports have noticeably longer range than other APs I've tested.  But there are several more to go.

2013-07-15 »

"You want to talk to Avery? ... Really? ... Oh, it's just that when people call for Avery, they're usually not so... happy about it."

– actually said by someone helpfully answering my phone

2013-07-16 »

Things I never thought I'd hear a PM say, Volume 1:

"I am opposed to pre-announcing availability. I'd prefer we seed the market by cherry-picking early adopters [...], then do a public launch where we simultaneously announce these early folks as success stories."

2013-07-17 »

In order to serve you better, I have renamed "wifi roaming" to "wifurcation."  You're welcome.

2013-07-18 »

#TIL  That the "Wi" in WiMAX has nothing to do with wireless.  The "MAX" does though.

2013-07-19 »

Best thing about iOS calendar notifications: they go away automatically a while after the meeting is over.

Every now and then I pick up my Nexus 7 and I have a zillion multi-day-old calendar notifications and way-obsolete GTalk notifications.  Cluttery.

2013-07-28 »

"You may avoid the forced logout by logging out now."

June 2013
August 2013

I'm CEO at Tailscale, where we make network problems disappear.

Why would you follow me on twitter? Use RSS.

apenwarr on gmail.com