Many facts are generally accepted, and provably wrong. I tend to be immediately suspicious when I hear a 'proof by popularity' argument. The entire body of scientific knowledge has been summed up by a wag as "Ideas that everyone knew were wrong, right up to the point where everyone knew they were right."

In this case I am trying to say that the fact is "right, with infinite resources". Since you can't have infinite resources, it's not quite right.

What the digital data represents is entirely irrelevant In fact, erroneously taking it into account has badly biased your argument.

You brought up the contents of the image in a previous post when you made an example of satellite images. It's not nice to switch horses midstream. Or are you trying to ride both horses at the same time?

The subject matters because any measurement, including a photo, has a margin of error. CCDs smear borders and have non-linear responses to hues and luminence, just like film. When the height of a mountain is calculated from imagery, it has a error of plus or minus x meters, where x depends on the measuring device. It is quite reasonable to loose data in this case, provided that x doesn't get too big.

Just as you can talk about the properties of a car's tires without mentioning the engine, you can talk about the properties of compression algorithms without discussing the camera that uses them.

Of course this is true - we certainly can talk about the properties of the tire without considering anything else. And yet, if you don't consider the engine, you run the risk of having the tire malfunction.

If you want to continue your tire analogy, this is like an engineer insisting on using a brand of rubber called 'superior rubber' because it has better shock absorbtion, while the other engineers attempt to tell him that tire will disintegrate when an engine is connected to it because the rubber is too soft. The engineer's main argument is "It's superior. Everyone else in the world understands the word superior. Why can't you?"

PNG itself is lossless, just like transmitting the original data is lossless. But in a practical application with limited bandwidth, tradeoffs always have to made, and the surrounding system must be considered. If we could ignore the surrounding system, there would be no need for compression - we could just assume infinitely fast machines with infinite memory and always work with the uncompressed original.

____________________
Jeremy
I didn't believe in evil until I dated it.

I tend to be immediately suspicious when I hear a 'proof by popularity' argument.

What do you think of defining terminology by popular agreement? And what do you think of "proof by definition?"

I'm not arguing merlyn's "point", jepri¹. I suggested that his use of the term "lossy", in a way which isn't only different from but directly contradicts the accepted terminology, would only cause confusion and wouldn't help anyone understand the issues.

The PNG and JPEG specifications themselves use the the terms "lossy" and "lossless." The PNG spec defines them:

Lossless compression
Any method of data compression that guarantees the original data can be reconstructed exactly, bit-for-bit.
Lossy compression
Any method of data compression that reconstructs the original data approximately, rather than exactly.

We don't need to argue anything, we only need to agree on terminology.

1. Really, his point was obvious. JPEG has its own advantages and they are different from PNG's.

-sauoq
"My two cents aren't worth a dime.";

Well, sure. But I got to accuse you of changing horses in midstream, which is something I haven't had a chance to do since high school debating ;)

The terminology thing is genuinely interesting though. You are pointing to a computer (science) definition, while I immediately reverted to my physics training. Both are technically correct, only the scope varies. The position you take would be supported be almost any mathematician, but probably few physicists.

____________________
Jeremy
I didn't believe in evil until I dated it.