Rastas!

Friday, November 7, 1997 by Dave Winer.

Earlier this week I was working on Windows, and at the end of the week I still am. But this morning I'm writing, and right now I prefer to write on my Mac. The words flow faster. The fonts are bigger, the screen is easier to read. And some of my favorite writing tools are here on my Mac.

I wanted to write this on Windows, but I have limited time, and I want to flow my ideas out quickly and get on to the next thing. I guess I'm a cross-platform kind of guy? We'll see!

Rastas! 

I'm thinking about rastas today because my Jamaican uncle and auntie just came for a short visit on their way home from Hong Kong via Thailand, India and Europe. I've been tinkin about the beach while I've been doing my geek work this week. A little patois in California. Ye-hi!

So, back to the beach. In Jamaica there are beach people called rastamen. They have wild hair and wild manners. They're funky indigenous people. They can also be a pain in the ass.

Walking down the beach in Negril, a resort community in Jamaica, the tourists like me, avert their eyes. It's about as friendly a place as most airports. This is weird because Jamaicans are very friendly people!

I have a theory why the beach in Negril is such a cold place. As you walk down the beach, a pair of rastas approaches. Yah man! Gimme 5. How are you me brother?

If you talk to them, two or three sentences later they will try to sell you something. A trip down a river to a waterfall. Or ganga, the local marijuana, or sex with their girlfriend; or maybe some fruit or meat. Sell sell sell. They have something they want you to buy.

The beach rastas in Jamaica are hustlers, they're a little charming at first, until you realize that they aren't really being friendly, they just want to sell you something.

Back to software! 

On Thursday, Pendragon Software, a company that develops benchmark programs for Java, ran a press release saying that Sun had done something weird to boost their numbers on one of their tests.

http://www.pendragon-software.com/pr1197-2.html

According to the press release, the Solaris implementation of Java seemed to be fifty percent faster at running one test than any other implementation of Java. Sun ran a press release that made this claim.

Pendragon was suspicious. They ran the test and confirmed the result. Then they made a small insignificant change to the test, and guess what -- Sun's Java wasn't any faster than any other Java.

What conclusion are we supposed to draw? That they have some secret sauce at Sun that makes their Java run faster? Could it be true? Well, no. In the real world, Java apps won't run any faster on Sun hardware than on others, even if the tests say they will.

Fudge 

Here's how Sun fudged the test. A user asks Java to do something. Instead of just doing it, Sun's Java asks who's asking. If the answer is Pendragon, they don't bother doing the work, they just return the correct answer. It's not surprising that they can do this fifty percent faster than if they had actually done the work.

Surely the tests could be more rigorous, I'm sure they will be, but here's another point of view: Sun could be honest.

Analogy 

Here's a reasonably close analogy from education. Imagine a teacher asking a student to do a homework assignment. The student wants to get the assignment finished as quickly as possible, and doesn't care about being honest.

The teacher doesn't really care what the answer is, presumably the teacher could do the calculations without asking the student. The teacher wants the student to do the work. The student wants to get a good grade without doing the work.

The importance of testing 

As we advance technologically, we depend on testing to keep us safe.

New drugs have to go thru long expensive tedious trials. Cars have to be crash-tested. We don't tolerate fraud in product testing in drugs and cars, should we tolerate it in software? No we shouldn't, if only because someday a car or a drug might rely on software as part of a testing process. Or a moon mission, or your tax return.

Computers are everywhere. We need to trust the people who make computers to have a high level of integrity. We want meaningful comparisons, and if a bit of software is slow or unreliable, we want to know before we bet our money or careers or safety on it.

Caught in the act, Sun says this is just the way business works.

They want to run the benchmarks well, they say.

Right on, I say, but do it honestly.

Dave Winer