The combination of “seems impossible” and “strong network effects” is about as close as you can get to the magic formula for incredible, sustainable success, as with eBay, Wikipedia, and Google.

Trenchant analysis from Joel Spolsky.

Sample fact: The time American’s spend watching television commercials every weekend is approximately equal to the entire time spent in creating Wikipedia to date.

Rule of thumb: “Doing anything is better than doing nothing.” “It’s better to do something than to do nothing.”

Via Jason Sutter who lives in New Zealand but shows up in my feed of blogs local to my zip code.

Update - 28 April 2008 Here (via Gruber) is a link to the text of the talk. Also, the quoted rule of thumb is corrected.

The days of waiting and worrying about the older girl as she lay ill in intensive care, twice, are also permanently engraved on my memory and stored in my body. I find and feel that concern every time that I get to see her and admire the young woman before me. That it all turned out well in the end was fortunately, but I still feel the toll.

I had two different concrete experiences of problem solving in the past couple of days. I think these examples are interesting because they are so contained, not because they are hard problems. They illustrate how I tackle a problem and provide some generalizable strategies for problem solving. I’ll cover the other problem in a future post and try to extract the strategies.

This post is about the little man’s homework this week. He’s still struggling with this:

Shelley wanted to buy some clothes that were on sale.

Blouses	Pants	Socks	Shoes
$10	$10	$5	$20
$15	$20	$10	$30
$25	$35	$15	$40

Her mother said she could spend $85. List all the different ways that Shelley can buy 1 blouse, 1 pair of pants, 1 pair of socks, and 1 pair of shoes that total exactly $85. One way has been done for you.

Blouses	Pants	Socks	Shoes
$10	$20	$15	$40
___	___	___	___
___	___	___	___
___	___	___	___
___	___	___	___
___	___	___	___
___	___	___	___
___	___	___	___
___	___	___	___

So how do you solve this?

The exhaustive solution is to run through each of the possible combinations of each item and total the associated cost. There are 3 possibilities for each of the 4 different items, so there are 3^4, or 81 possibilities. That would be a moderately tedious number of sums for an adult faced with a real life problem. The little man is 8 and doesn’t care about clothes, so he has zero patience for that approach. He wants to look at the problem and figure out the solution and if he doesn’t get it immediately then he wants to write, “I tried my best, but couldn’t figure out the answer” on his paper. Apparently this is some sort of magic incantation that he thinks his teacher will accept.

His parents aren’t happy with that solution. He has only investigated approximately 2 of the 81 possibilities, the top row and the given solution. I don’t know how to motivate him to check out all 81. I still don’t. That’s not what this post is about, but if you have any ideas, let me know.

I got bored myself after running through the first 9 possibilities — there are no solutions with the $10 blouse and the $10 pants — and started looking to solve an easier, less compute intensive, problem.

I took a couple of approaches.

First thing that I noticed was that there had to be either 1 or 3 items whose price ended in 5, or you couldn’t get $85. As I was staring at the possibilities, I felt like this eliminated many options. Many is a subjective word, though and after I figured out the there were still 39 remaining, it didn’t seem so many. Computing 39 sums of 4 two-digit numbers is tedious, even if all the terms are multiples of 5.

Human calculation speed is greatly increased by dividing every items price by 5, and trying to get them to total 17 (=85/5). You can still limit this to possible solutions that have either 1 or 3 odd terms.

None of the shortcut solutions are ultimately satisfying however, because of a flaw in the display of the problem. Even running through all 81 permutations won’t satisfy, because there are spaces for 9 solutions, but there are only 8!

I wrote a little program that iterated through all the possible solutions and found these solutions:

Blouses	Pants	Socks	Shoes
$10	$20	$15	$40
$10	$35	$10	$30
$15	$20	$10	$40
$15	$35	$5	$30
$15	$35	$15	$20
$25	$10	$10	$40
$25	$20	$10	$30
$25	$35	$5	$20

Ultimately, this becomes a problem of combinatorial optimization, which seems like too much for 3rd grade.

Anyway, I haven’t shared my results with him. I don’t know if I will. He might be fascinated, or repelled, by them. Most likely, he would want me to give them to him so he could copy me and just get it done.

Thanks Rodgers.

When it came time to pay the ticket, I elected to pay an additional $32 to the City and County of San Francisco and $20 to online provider of traffic school in this jurisdiction. This allows take a course and a test. Once I pass the test, I can have the infraction removed from my driving record, protecting my insurance rates, and presumably rendering me a better, safer driver.

I went through the course materials and took the chapter tests. The content was reasonably well presented, although the web site was annoyingly slow and there were a number of infelicitous UI decisions (arrows that seemed to imply links, but didn’t, no url or other identifying information in the e-mail that they sent me to confirm my registration, etc.)

Some of the material that was presented in the course and the chapter quizzes didn’t seem like it was particularly important to the goal of improving my driving skills and making me a safer, more responsible driver. Material on what constituted valid forms of identification at the DMV is useful, but it isn’t going to keep me from pulling out into traffic when I don’t have the right of way.

Little did I know. After I completed the course, I was directed to take a final exam that was administered by NTSA, the National Traffic Safety Administration, which is not the NHTSA, the National Highway Traffic Safety Administration. NHTSA is a branch of the Federal government. NTSA is not:

NTSA provides home study traffic school course review and evaluation to insure compliance with all Court requirements and standards. All home study traffic school course reviews conducted by NTSA for court approval, are conducted by Mr. Ernie Garcia, a retired Highway Patrol Lieutenant with 30 years of experience.

In theory, NTSA is an independent agent that can ensure that the student has taken the course and has learned the material. Each independent provider of traffic school sends their students on to NTSA for final testing.

This final exam is crazy. I thought that my browser had been hijacked, or that I was the subject of a prank. Some of the questions were ridiculous beyond belief.

Examples:

In case you aren’t following me, these three questions pertain not to the Motor Vehicle Code or rules for defensive driving, but to the very specific version of the course materials that I studied. The first refers to a specific illustration in the course materials, the second and third to how the material was presented.

It gets worse.

This test question is asking me about the background color of a particular item?! Unbelievable.

Worse, there is no right answer listed. The background is neither yellow, purple, orange, red, nor blue. It is white, I checked the html.

The background color for that particular paragraph is the same as the preceding one, only the font color is changed to red. Red is not the correct answer.

There are 40 questions on each test. The student must get 20 out of 24 on the questions that pertain to DMV materials and 13 out of the 16 questions that “NTSA develops with each traffic school, security, or course specific questions that relate to specific events in their course that have nothing to do with the required curriculum information of the course.”

Is anyone well served by these questions? The point is to ensure that the person taking the exam is the same as the person who took the course. Or, or from the exam web site, “To reasonably insure that the person taking the traffic school course is the one taking the final examination.” I don’t suppose that you could accurately answer these questions unless you had looked at the course, but I read through the course and I couldn’t remember how many rows of signs there were in a particular illustration or what the first illustration in Chapter 3 was. I have a reasonably good memory for this sort of material, but come on.

Couldn’t the independent traffic schools content be differentiated by content instead of layout? The course materials had various random content, jokes, even a couple of lines about how the school website was created with Dreamweaver. Do they really have to test on formatting?

I was expecting the blood and gore movies of my high school driver’s education classes. This is scarier.

It just reinforces my previous beliefs that traffic school is a scam.

Drive safely.

For those who don’t know, MarsEdit is a Macintosh program to write blog posts.

There are a couple of small features in the way the Mars Edit works that could have been deal breakers for me, so I was going back to the MarsEdit forums to figure out if there were a work around for them. I couldn’t remember the url of the site, so I Google MarsEdit.

The problem is the results. The first result returned by Google is to a page on the Ranchero Software site mentioning that Ranchero Software sold MarsEdit to Red Sweater Software more than 7 months ago. The second Google result points to the Red Sweater Software page that describes MarsEdit.

Curiously, the PageRank for the Red Sweater page is 7 while the Ranchero page PageRank is 0, and the PageRank for the Ranchero Software home page is 7, while the Red Sweater home page is 6.

What this means is that there are so many more links out there on the web that point to MarsEdit at the Ranchero site than at the Red Sweater site, and the rating of those sites is so much higher, that a deficiency in page rank of 0 to 7 is overcome.

Even more confusingly, the absolute number of links isn’t that different. As of this moment 282 links point to the Red Sweater page and only about 592 links point to the Ranchero Software site.

I wouldn’t have thought it possible. This is a PageRank result that really confounds all my expectations.

Her version of the story:

R: I thought of something else you should get at the store.

L: Ok.

R. Hum.

L: I’m getting a pad.

L: What was it?

R: What was it?

L: What was what?

L: What you wanted to me to get?

R: I thought you wrote it down?

L: Wrote what down?

R: Gum

My side:

R: I thought of something else you should get at the store.

L: Ok.

R: Gum.

L: I’m writing it down.

L: What was it?

R: What was what?

L: What you wanted to me to get?

R: I thought you wrote it down?

L: Wrote what down?

R: Gum

Most frequently when I am searching, if I enter multiple terms I am implicitly requesting results for items that match all of the terms. This is the way that results from Google and Yahoo work. If you search for Repair Manual the results are pages that contain both the term repair and the term manual.¹

I’ve bolded and in the last sentence because that is the operative term. The results returned by Google and Yahoo are the logical conjunction (and) of the separate results for the individual terms. If I search for Repair and Manual, Google helpfully informs me, “The ‘AND’ operator is unnecessary — we include all search terms by default.”

The other commonly encountered method of combining results is disjunction. This search returns all pages that contain either repair or manual. There are obviously many more results in this case. For repair and manual Google returns not quite 2.5 M results, while for repair or manual Google returns more than 40 M results.

The problem that I encountered, was that many web sites silently ignore the de facto standard set by search engines and attempt disjuctive, or, searches. Instead of returning pages that contain repair and manual they return pages that contain repair or manual. I’m not exactly sure why so many e-commerce sites seem to do this, but it may have to do with available tools. I know that disjuctive search is the default for the Apache Lucerne search engine library, although Ferret, the Ruby search engine gem which was inspired by Lucerne, switches the default to conjunctive search.

Can you see where this is going?

I was attempting to repair a dishwasher that had stopped cleaning the dishes on the top rack. Some small piece of plastic - perhaps the inner liner of a yogurt container - had gotten loose in the dishwasher, been sucked into the pump and macerated. Small pieces of plastic had been blown through the entire circulatory system. I had cleaned much of it out, but it still wasn’t working and I wanted a closer look, but I wasn’t sure how to take some parts off without breaking them. All I needed was a Repair Manual for my washer.

A little Googling around led me to an appliance repair site. I located my make and model of dishwasher and searched for repair manual. When there was only one result returned, I wasn’t surprised. I only expected one.

Without any closer examination I added it to my cart along with the other part that I knew I needed.

Unfortunately, when the part arrived, it was accompanied not by a detailed Repair Manual but by the ludicrous User Manual that I already had.

I was a little annoyed with myself for not being more careful and a little annoyed at the merchant for giving me a misleading answer to my question, but the cost was minimal (perhaps I would have been more careful if it had been higher).

As I noted above, this experience has recurred several times since, minus the steps of adding an incorrect item to my shopping cart and buying it. I think merchants are making a mistake in ignoring the search engine standard and courting a possible liability.

Customers are dissatisfied by these unexpected results and if they make a purchase based on them, they are going to want their money back.

1. Both search engines do some other processing, so the returned pages might contain “manuals” instead of “manual” or even “fix-it” instead of “repair.”

If you do try to poke around on the Internet looking for prints, you might just come across this reproduction of the Great Wave (local cache). This is a fine lovely print, you can zoom in to see the detail, read a brief description and note the bequest by the Havemeyer family. (Zoom in on Mt. Fuji and the boats to appreciate the phenomenal detail.)

Art lesson over, the thing that jarred me about this web page was clicking on the “Next” button. That goes to this page (cached) containing a pair of Royal Earrings from 1st century Andhra Pradesh, India.

Could anything be more miscellaneous? Different styles, different techniques, different media; two works of art separated by two thousand years and four thousand miles. They have nothing in common beyond being art produced by human beings. To be fair to the Met, I jumped into a spot in their collection stream where it transitioned from Japanese to South-Asian, so I accentuated the effect. It was as if I were present in the gallery and I turned the corner from the East Asian Hall into the South Asian Hall. On the web, I didn’t have any visual clue that I was nearing a transition. The collection streams are organized chronologically by region, further heightening the transition.