Slashdot

Recently I used my limited artistic skills to produce a cartoon of a giraffe's failed suicide attempt. In a flight of fancy, I thought about how I might sell it as fine art rather than mass merchandising it. AFAIK, one of the selling points for certain works of art is that you're buying the original.

But with digitally created pieces, how can you sell the original? How could you certify one digital file as an original vs. a copy? Would you sell the hard drive you stored it on while creating it? Would you buy a computer for each digital work and sell the computer? Do you sell the copyright with the piece, so you surrender the right to make copies?

Part of the value of art sold to collectors is in its scarcity, either through never copying it or allowing only a very few copies in a limited edition. How would you create that sense of uniqueness so that you can add that scarcity premium when selling a digital work?

What's the best way to measure a CGI script's usage of system resources (processor cycles, RAM) to predict how many users you can handle at once, or within a short time period, before your server is "slashdotted"? I figure there has to be a utility for stress testing a CGI-script or even a plain web page and measuring what levels of use start degrading performance and to what degree. So, any recommendations?

So I have a piece on my blog about detecting mobile browsers and some PHP code to do it. One part of the code checks 67 text snippets against the User Agent string, and a visitor asked if the way I was doing it -- putting all the snippets in an array and regex matching them one at a time -- was more efficient than using one giant regex with all the snippets in it.

So I actually tested it. When a matching pattern was hit very early on in the text, the giant regex was faster. But if there was no match or the match was much deeper in the text, the batch of smaller regexes performed better. Essentially, the batch of smaller regexes took the same time to run whether or not there was a match or where it occurred in the examined text. The giant regex slowed down the farther the match happened in the examined text and was slowest when there was no match.

As the length of the examined text grew, particularly with no match, the disparity between the giant regex and the batch of regexes grew. At 760 characters and no matching text, the giant regex was taking around 4.25 times longer than the batch of regexes. At 1520 characters (around 250 words), the giant regex had slipped to 5.4 times longer. Yet if there was a match in the first 10 characters of the 1520-character string, the giant regex was the same speed as if the match had come in the first 10 characters of a string 1/8th the size.

So while the giant regex would be more computationally efficient against shorter strings or ones where you knew the match would come early, the batch of many smaller regexes is actually better as a rule of thumb. It's faster against larger blocks of text when the match is deeper or non-existent, and if the text grows, it's computational needs don't grow as fast as those of the giant regex.

In Friday's edition of "Dear Margo" (an advice column written by the daughter of Ann Landers), a reader asked if Yahoo! employees were reading her Yahoo! mail. Margo addressed the question with a quote from her boss, a Yahoo! News VP. He not only answered the question with doublespeak, but used it as an opportunity to disparage Google's Gmail.

Many web site owners use Google AdSense to help cover the costs of their sites or earn some money from them. But for nearly a week now, Google's reporting of exposures and clicks has been at least partially broken. I've been tracking this on my blog since I noticed one of my ad channels was reporting around 7% of what it should despite my site traffic being at normal.

Google acknowledged there were discrepancies after it became so bad you couldn't help but notice it, but they only did so on a thread in the Google Groups AdSense Troubleshooting group and have yet to offer an explanation or ETA for a fix. The next day numbers went down again, this time by a factor of 10 or so, and Google remained mum. As they started coming back up on Thursday, Google chimed back in with another "we're working on it" and that was it.

They say that this problem is just affecting specialized tracking, but isn't affecting aggregate reporting for a whole domain (basically, you're making all the money you should, but you just can't track where it's coming from in as much detail), but many webmasters are complaining of lower revenues to go with the lowered channel numbers. I know my clickthrough rate was the lowest I'd seen in months on the same day this problem was at its worst, though that could be a coincidence. Despite this, Google remains tight-lipped, giving webmasters a minimum of information and providing a minimum of reassurance (two posts in three days that merely state they know it's happening and are working on it). Hopefully they'll be more forthcoming when this is solved and they can do a postmortem on it. But, in the meantime, they're making a lot of people nervous.