Bryce Kerley

You can't use prime numbers to cheat on CRDTs

Tue, 10 Apr 2012 20:50:16 -0400

Conflict-free Replicated Data Types, or CRDTs, are a simple way to make a data structure that works well in an eventually consistent environment, such as the Riak distributed database my employer makes. I’ve had “google CRDTs” on my to-do list for months; each time I had a hard time getting in to it. Last week, somebody linked me to Kyle Kingsbury’s Ruby implementation of them, and I was very very happy to see that the README file is awesome.

Seriously: https://github.com/aphyr/meangirls has an awesome README, easily the best CRDT resource I’ve read. Thanks, Kyle and Sean Cribbs!

Counters and Primary Keys

One thing that you get with a consistent database (such as Postgres, MySQL, and Redis) is a counter that you can use to generate nice-looking IDs that are not enormous. Think the “943” in https://github.com/blog/943-rubyconf-block-party, vs. the “182302468733018112” in https://twitter.com/#!/HIMANSHU/status/182302468733018112.

You can use a G-Counter (see the meangirls README) to count things, but it doesn’t work as a primary key: if two nodes increment their entries at the same time, they’ll both have the same value.

{
  'type': 'g-counter',
  'e': {
    'a': 1,
    'b': 5,
    'c': 2
  }
}

If a and b increment at the same time, they’ll both get a new value of (1+5+2) + 1 = 9. Since the whole point of a primary key is to be unique, that won’t do.

Prime numbers

Now, what if each node has a different prime number that they increment by? Let’s make nodes a, b, and c use 3, 5, and 7, respectively.

{
  'type': 'g-counter',
  'e': {
    'a': 3,
    'b': 25,
    'c': 14
  }
}

This counter has a value of (3+25+14) = 42.

Now, if a and b increment at the same time, a will see (3+25+14) + 3 = 45, b will see (3+25+14) + 5 = 47. When c looks later and merges them, it’ll come up with 49, but that’s not a problem: we just didn’t want a and b to increment to the same value.

Here’s the problem: if a increments 5 times in the same duration that b increments 3 times without their versions getting merged, they’ll land on the same number, since 5 * 3 = 3 * 5.

Big prime numbers

You can use prime numbers larger than 3 and 5 to mitigate this a bit. I’ll be using the RSA key generator from http://ats.oka.nu/titaniumcore/js/crypto/RSA.sample1.html set to 16 bit keys for this.

Here are two bigger primes: a = 185399 and b = 125729 . You’d have to increment a 125729 times and b 185399 times without a merge to run into a conflict. Let’s see the friendly numbers from the previous example ( c = 200819 ):


{
  'type': 'g-counter',
  'e': {
    'a': 185399,
    'b': 628645,
    'c': 401638
  }
}

This counter has a total of… 1215682 . Not quite a friendly 943 , but not as hostile as FD22C314836F11E19425AB7961E300D6 either.

My Top 5 Artists (Week Ending 2011-12-11)

Thu, 15 Dec 2011 23:09:02 -0500

My Top 5 Artists (Week Ending 2011-12-11):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-12-4)

Mon, 12 Dec 2011 03:50:19 -0500

My Top 5 Artists (Week Ending 2011-12-4):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-11-27)

Tue, 29 Nov 2011 08:58:22 -0500

My Top 5 Artists (Week Ending 2011-11-27):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-11-20)

Tue, 22 Nov 2011 14:12:59 -0500

My Top 5 Artists (Week Ending 2011-11-20):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-11-13)

Wed, 16 Nov 2011 08:41:05 -0500

My Top 5 Artists (Week Ending 2011-11-13):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-11-6)

Wed, 09 Nov 2011 03:39:11 -0500

My Top 5 Artists (Week Ending 2011-11-6):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-10-30)

Sat, 05 Nov 2011 22:32:28 -0400

My Top 5 Artists (Week Ending 2011-10-30):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-10-16)

Tue, 18 Oct 2011 00:22:56 -0400

My Top 5 Artists (Week Ending 2011-10-16):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-10-9)

Fri, 14 Oct 2011 15:06:38 -0400

My Top 5 Artists (Week Ending 2011-10-9):

Imported from Last.fm Tumblr by JoeLaz

Photo

Mon, 12 Sep 2011 22:19:56 -0400

globochem: Anamanaguchi - Rainbow in the Dark (Das Racist...

Sun, 28 Aug 2011 23:14:25 -0400

globochem:

Anamanaguchi - Rainbow in the Dark (Das Racist Remix)

i’m in the building, building with belding

somethin like a neo-rap zack attack

My Top 5 Artists (Week Ending 2011-8-21)

Thu, 25 Aug 2011 02:41:27 -0400

My Top 5 Artists (Week Ending 2011-8-21):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-8-14)

Sat, 20 Aug 2011 02:07:31 -0400

My Top 5 Artists (Week Ending 2011-8-14):

Imported from Last.fm Tumblr by JoeLaz

Car Review: 2011 Chevrolet Cruze

Wed, 13 Jul 2011 18:35:00 -0400

The cars used in this review were provided by General Motors.

The Chevrolet Cruze is what I would consider to be an “compact car;” it’s not frighteningly big, not frighteningly small, good on gas, and is the kind of car I would drive to work in if I had a commute.

Most of my impressions are from a weekend where I drove from Miami to Melbourne, three hours each way. The trip went like this:

Use switchblade key to open trunk.
Load trunk with bags, notice that there’s slightly more room than in the trunk of my 1997 Mitsubishi Mirage, much more room than the 2011 Chrysler 200 Convertible or the 2009 Chevrolet Corvette Convertible, and much less room than the Ford Crown Victoria Police Interceptor used for taxis in Miami.
Close trunk.
Unlock car, sit down, and buckle up (always buckle your seat belt when you enter a car or truck).
Plug iPhone 3GS into armrest for charging and music.
Start car, drive away.

The driving experience of the “LTZ RS” version I had was pretty standard. It didn’t feel floaty, shaky, or otherwise frightening at highway speeds in the least; but at the same time it didn’t have a ton of power to spare for passing. I suspect I’d enjoy the manual transmission version more on this front, since the auto is slow, even when manually commanded (although it does have six gears). Coming off of a stop, there’s a sweet spot above 3500 rpm where you can feel the extra power from the turbo, but it’s not a sports car by any means. This also means that it’s not loud and that the suspension isn’t hair-raising and jittery, which is nice.

The interior is quite nice for the size of the car; it’s better than the 2010 Prius, but not the 2009 C-klasse. The fabric covering on the dashboard looks nice, and doesn’t feel as hot as a plastic dashboard does after a day in the sun. The leather seats were okay, although I think fabric ones would be nicer in Florida. The instruments are easy to read, but I think the VFD display between the gauges could stand to show a bit more information.

In particular, while the display does give you some nice options such as the next navigation item, average fuel economy, and range until empty, it could stand to maybe put two things on the same page, provide an instantaneous fuel economy readout, and maybe a bit of audio information.

The OnStar navigation works, but it’s a bit janky and slow to get it set up. It’s nice that the car doesn’t store any map data that will be out of date long before it’s actually loaded into the car, but calling OnStar and using voice recognition to load in a new route, and being unable to quickly recover from a missed turn is pretty lame compared to what you see on a cell phone or other cars with navigation.

The audio system was pretty reasonable; it sounded fine, the XM worked well, but browsing a large collection (> 100 albums) on my iPhone was hard, since there’s no voice search, just a four-line screen and a dial to page through it. Dialing the phone over Bluetooth was also a dicey proposition; I never was able to call somebody by name (including difficult entries such as “Home” or “Shawn Baker,”) and had to use the dial to pick from the whole phone book or recent call lists pretty often. My #1 annoyance was that the navigation system wouldn’t duck out in a phone call like the stereo would, so everyone got an earful of turn-by-turn directions.

Since it’s a compact car, I didn’t try and put four 6’ guys in it for a trip to Epcot; my parents got the vehicle for the day, and liked it. My dad in particular said that he’d consider one once the reliability is a known factor, and that him and my mom didn’t have any really problems with it around town.

I didn’t see fantastic gas mileage during my travels, but did get about 35mpg on the highway with A/C on. Other than that, there’s not much to say; it’s a quietly competent compact car, with nice toys inside.

My Top 5 Artists (Week Ending 2011-6-26)

Thu, 30 Jun 2011 10:28:08 -0400

My Top 5 Artists (Week Ending 2011-6-26):

Imported from Last.fm Tumblr by JoeLaz

Short BusyCal Review

Tue, 21 Jun 2011 19:00:53 -0400

iCal annoyed me with frequent error messages about Google’s janky-ass calendar service being occasionally unreachable. On occasion, these messages would block the UI in such a way that required me to force-quit iCal.

BusyCal doesn’t! I’m willing to overlook some UI strangeness (goofy panels, fonts that aren’t pixel-perfect to iCal) for this.

Try it for 30 days perhaps!

This review is based on my experiences reviewing BusyCal during the 30-day trial period. I have not received any compensation, and nobody affiliated with BusyMac contacted me prior to publication.

My Top 5 Artists (Week Ending 2011-6-5)

Wed, 08 Jun 2011 04:16:59 -0400

My Top 5 Artists (Week Ending 2011-6-5):

Imported from Last.fm Tumblr by JoeLaz

My Top 5 Artists (Week Ending 2011-5-29)

Sat, 04 Jun 2011 21:15:52 -0400

My Top 5 Artists (Week Ending 2011-5-29):

Imported from Last.fm Tumblr by JoeLaz

New Jargon for Gadget Blogs

Mon, 30 May 2011 12:39:41 -0400

I read a troubling article today suggesting that a plastic mock-up containing a printed-on screen and no electronics was a product. With that in mind, let’s clarify what wording should be used in this situation:

“make-believe:” a thing that doesn’t work, has a simulated screen, or contains no electronics.
“pretended:” unveiling a make-believe product
“playing with toys:” hands-on “usage” of a make-believe product

For example:

Wrong:

We just got to spend some quality time with ASUS’ newly unveiled Padfone and you can see the smartphone that doubles up as a tablet on video after the break.

Right:

We just got to have playtime with ASUS’s newly pretended Padfone and you can see the make-believe smartphone that doubles up as a make-believe tablet on video of us playing with plastic toys after the break.

Thanks for your cooperation.