Feedback and Comments

In September 2005’s Linux Magazine, a reader asked why magic numbers are a better solution than Windows’ s file extension mechanism. I wanted to add to[ “Power Tools” columnist Jerry Peek’s] reply.

I can think of several reasons that magic numbers are better than extensions.

First and foremost, extensions on Windows are little more than vanity. Microsoft got to .doc first, so they own it. If I want a document format, I have to name it something else. In addition, because there’s no (global) registry, my nifty extension loses as soon as some big vendor steps on me.

For the use of short names, Apple did much better with file types. The Macintosh keeps the type (or kept, as I don’t know how this has evolved), the equivalent of the extension for this purpose, in a consistent location and vendors have an opportunity to register new types. Unfortunately, dealing with a centralized authority of “magic numbers” is also problematic.

Another concern about extensions is that there are so few of them. I expect modern versions of Windows allow extensions longer than three letters (but does the extension-to-type mapping system allow it, I don’t know) and allows extensions containing any Unicode character, but in practice, we’re talking three monocase, alpha-numeric characters. That’s less than 50,000 possible file types (mathematically), nevermind the fact that no one wants the .sux extension. Lest 50,000 seem like more than anyone could ever need, I observe that there are roughly 10,000 entries in the magic numbers file on my Windows machine, so a fifth has already been consumed.

The file name and a description of what the file contains should be independent.

In the August 2005 feature “Git With It” by Sam Williams, Mr. Williams refers to Linux as a “… reverse engineered version of Minix. ” This is not factually accurate.

Linus Torvalds has said several times that he used the system call descriptions in documentation from the Sun operating system to create Linux. Mr. Williams must be a Linux novice not to be aware of the very public and difficult dispute between the creator of Minix and Mr. Torvalds.

I read a lot of “tech” magazines, everything from MacTech to SysAdmin. Rarely does one come across good writing in these magazines, at least not as good as Sam Williams’s piece on Git.

One might argue that the audience for these types of magazines cares very little about good journalism; instead, the readers want to find out how to get bash to play MP3 files or how to download Debian on their cell phone. Yet I suspect those who are interested in complex issues like free software development tools are also appreciative of a well told story, and Linux Magazine is doing the right thing by publishing articles like “Git With It.”

I, for one, care a great deal about good journalism and am grateful for the high standards your magazine holds to, especially compared to its highly technical, yet slightly convoluted brethren.

Thank you for the nice article on Zoidberg (available online at http://www.linux-mag.com/2005-09/power_01.html). The article highlights a few features I’m proud of and motivates me to get zoid out of beta into stable.

There’s a silly statement in the October 2005 feature” Opening Windows”. The author writes:”[…] such offerings are a long way from publishing[…] the well-guarded application programming interface specifications that allow a third-party program to run seamlessly atop Windows”.

Hasn’t the author heard of MSDN? Has the author ever visited http://msdn.microsoft.com? How exactly does he think third-party application developers write Windows programs?

I’ve been receiving Linux Magazine in prison for the last several years and I’m being released August 1,[ 2005]. I am including a change-of-address request.

I managed to thwart the prison rule of no more than five publications in my locker by hole-punching the issues and putting them into one large binder, thereby counting it as one “book.” Plus, I’ve had some really decent guards. Keep up the good work!