Slashdot Mirror

I think it was a major strategic blunder not to do the IPO last time around. I like Google as much as the next guy, but even the most loyal Google supporters have to admit that their search results pages (SERPS) are now filled with spam. Half the pages found with high-profit keywords are total fluff and pages crafted by Search Engine Optimizers to grab as much Google traffic as possible without providing any real content. Google's algorithm, (and its reliance on inbound links) has been reverse engineered to a point where it is no longer valuable for the most coveted (and high-competition) keywords.

So they need to do their IPO as soon as possible, as there's more competition coming down the pike... and Google's place in the universe is far from secure. To draw a gambling analogy, it's time for these guys to cash in their chips. I'm not saying that they are going to go away (or even that they will lose the war) but there will very likely never be another time where their company's name is on the tip of the tongue of every American, and where their company is held in such high regard (which provides a perfect environment for a successful IPO.)

I'm rooting for them, but if their SERPS don't get cleaned up soon I'll be taking a serious look at their competition. I doubt I'm alone.

--- JRJ

In related news, a recent study found that the average computer user is an idiot. Film at 11.

That's not fair, of course. For example, try searching for spyware removal software like "Spybot Search and Destroy." Almost all the links you'll find are for imposters that are themselves spyware. Evil.

Earthlink has their own spyware removal sofware, but I'm amazed it doesn't get caught in an infinite loop installing and removing itself, since Earthlink's software includes spyware.

--- JRJ

Of course your concept assumes that Microsoft has actually invented something. (Cheap shot, couldn't resist)

--- JRJ

Oooh! Oooh! It's the year that Linux is finally going to take over the desktop... again. Just like 1997 was. And 1998. Oh, and 1999. 2000? 2001? 2002? 2003? Sensing a trend?

As Bill Gates himself says, we often over-estimate the impact of a given technology will have in 5 years time, but we tend to UNDER-estimate its impact over 10 years. I think that the Linux on the desktop is similar: it will gain marketshare, but MUCH more slowly than people on /. (or even Linux-friendly journalists) assume.

Let's stop measuring progress in years, and start measuring it in decades-- only then will we see the impact that Free software is having. Revolutions take time.

Oh... and balls through windows? Could you have come up with a weaker punn? :)

--- JRJ

Don't get me started on how audio fingerprinting works. :)

--- JRJ

I've done extensive development work in the area of audio watermarking and audio fingerprinting, and I'm amazed that AT&T can make this happen, given the reduced fidelity of a wireless phone connection. Music fingerprinting technology is a smaller (and more approachable) problem domain than open-ended speech recognition, but still this is quite an achievement.

I congratulate them on the technical achievement, but I think that $0.99 (which is the price quoted in the review) is way too high a price for this service-- for that I could actually buy the song on iTunes or Napster. Unless they drop the price, I don't think this service will be terribly successful.

On an interesting note, it is not clear from their TOS whether or not you still have to pay for a song recognition even if the service is unable to accurately provide you with the song title.

Cool idea, but not for a buck.

--- JRJ

• CPU time is, indeed, CPU time. It is based on the amount of time the CPU actually spends on the given process ID. I cranked everything for 60 minutes, and the total CPU time was 41 seconds.
  
• It sends traffic on port 25 as expected.
  
• SMTP failure does not seem to have a short-term impact on the calculated CPU time. I ran it for 30 minutes where it was being tricked into thinking the messages were sent, and 30 minutes where it WASN'T being tricked into thinking the messages were sent. The CPU time was about the same for each of those half hour segments. (19 seconds for the first, 23 second for the second.) They don't seem to be checking. These numbers WERE reflected on the web site's management area as well.
  
• based on the above info (i.e. that they are not checking to see if the messages are actually being sent) I have to assume that they don't plan to pay anyone. I can't imagine they would have made that mistake otherwise.
  
• Their backend (which handles the requests) seems to be web services based, and doesn't seem to have any form of authentication in place.
  
• The app is pretty configurable. I've posted some informative screenshots on my blog.