Alston Grove

Steve Wozniak (who likes to be called “Woz”) of Apple fame is back in action. You may remember that after engineering the early Apple computers, Woz quit to teach public school. Now, 20 years later, he’s started his own company called Wheels of Zeus to make a low-cost GPS tracking system. The “wOz Platform” includes a reference design, wireless network, and online service that enable people to locate and take better care of what’s important to them. With venture funds from multiple firms and corporate commitments from Motorolla, there seems to be momentum, but there is no product available yet. An article at Wired talks about wOz.

I admit paper is the ultimate surface for reading text. But my collection of tech magazines is using up too much physical space. Wired, Linux Journal, Dr Dobbs, MIT Review, Business 2.0. Stacks and stacks of magazines clutter my shelves. I feel like a pack rat, so instead I just now put them in a recycle bin at the curb. But I feel bad, because what if I needed an old article for reference?

Weren’t these magazines in some electronic format in order to typeset and print them? Of course! Why didn’t my subscription earn me access to an electronic version? Imagine 5 years of a publication on CD or DVD. It could be indexed and searchable. And it would fit nicely on my basement file server. Linux Journal actually is available on CD-ROM, probably because they use SGML, XML, or TeX for their typesetting. Except, I have to pay another $30 for the CD after I’ve been a loyal subscriber? At least it’s available! Wired Magazine hits newstands and then main articles appear online shortly after. Why did I bother to subscribe? Well, it was only $5, and the paper magazine is pretty. But why can’t I get a year of the mag on a CD? As for the other mags, they suck.

Forgetting magazines for the moment, why aren’t more books in eBook or some online format? Amazon has a growing eBook section. But selection is paltry, especially tech books. O’Reilly does a much better job with their Safari Books Online. Unfortunately, the basic monthly subscription price of $14.99 needs to come down about $5.

Why are tech books so expensive anyway? If I want to buy a new fiction book like The Da Vinci Code, it costs me $15, but a book on Programming Java costs me $85. Come on! Do programmers make too much money, or are most tech books bought by corporations for training?

And then I compare the paperback version of tech books to the eBook version, and the price is the same or $5 cheaper. What?! We’re not procuring matierals and manufacturing bits here — it costs nothing to make a copy of bits and very little to pay for shared bandwidth. The publishing industry needs to get the cost low enough for the convenience of buying the PDF instead of finding it on BitTorrent for free. Before this medium becomes more popular, they better look at what happened with the music industry.

I think eBooks are about to become more popular as CRTs are displaced by LCDs. An even bigger innovation, called e-ink, is about to create buzz and provide an environment for eBooks to really take off. Sony’s e-Book reader LIBRIé is the first display using electronic paper ink. Now, if they don’t screw consumers too much with the DRM and the cost of eBook files, we can all be happy.

Is there a shift occurring from file sharing to stream ripping? Listening to streaming radio stations like those on Shoutcast doesn’t give you control over which songs you hear, but you can choose a genre and a high quality bitrate. And the RIAA doesn’t know if you are just listening or ripping the mp3 stream. Links from an article on Slashdot on the topic have people discussing the software they use and how the RIAA might respond.

I’ve been using Streamripper, which uses the ID3 data in the stream to write mp3 files with the right ID3 tags. Sometimes, the ID3 data arrives too early or late, and the files are not cut up cleanly, especially when the broadcaster is using cross-fading. Sometimes the stream hangs up after a while with certain stations, which surprised me over broadband. Are some broadcasters doing this on purpose to foil rippers?

Recording radio with a computer is protected by the Home Recording Act of 1992. Because computers are general purpose machines and not digital audio recording devices, they are not required to comply with Serial Copy Management System requirement (no copies of copies). However, Real Networks sued Streambox using the Digital Millennium Copyright Act to prevent them from saving streams from Real audio format. But if there is no copyright protection on the stream format (straight mp3), then is there no circumvention?

Another interesting idea is XM Radio. They sell a nifty USB-based device called XM Radio PCR that lets you listen to satellite radio on your computer, which means you can save the audio easily. It works using a line-in to your sound card, so the signal is analog. However, the PCR has already been hacked, and you can either buy a kit or a modified PCR to get digital optical outputs. Of course, there is XMPCR Linux software available, based on a Perl module, that makes ripping the stream and tagging the artist information easy to do.

I was experiencing random segmentation faults while running programs under Linux. I was encoding some OGG files and oggenc kept crashing from it. Suspicious, I tried compiling the kernel source, and sure enough, segmentation fault again.

Also known as signal 11, sigseg, or segfault, this signal from the kernel means that a program used a pointer to reference invalid memory. Usually it means a buggy program, but when it starts happening system-wide, there is more likely a hardware problem. Most programs will run without a problem, but code that uses a lot of computation and pointers (like encoders and compilers) are bound to get the signal quickly.

I immediately blamed my memory chip, a single 512M SDRAM. There are a couple of unique programs for Linux that help with bad memory. First, Memtest86 can run patterns across all memory and monitor for bad addresses. Next, the Bad RAM patch for the Linux kernel will allocate the bad memory addresses at boot time, making sure they are never used by running code.

But after running Memtest86 for hours, it didn’t find anything wrong with memory. Then I tried restricting Linux to only 64M (using the “mem=64″ kernel parameter), but I still got SIG11. I thought the odds of bad memory being in the first 64M seemed unlikely, so I started looking around for another source of the problem.

On the PC Health page in the BIOS, I noticed my processor temperature running at 105 degrees Celsius! Even an AMD Althon shouldn’t run that hot. After examining it on the motherboard, everything looked okay. But the fan seemed a little too dusty, so I cleaned it out. I pulled up the BIOS page again, and the processor had evened out at 73 C. I booted up Linux, ran some heavy processing, and SIG11 had disappeared! I’ll never underestimate the power of dust again.