Slashdot Mirror

This list is more comprehensive.

This isn't unheard of. To toot my own horn a little, my msort sort utility has an optional GUI with a "Show Command Line" command on the File menu.

I saw something about this a couple of weeks ago and blogged about it because the numbers seemed off. The cited chart, which is the same one I used, gives the volumetric coefficient of thermal expansion for gasoline at 20C as 950e-6, which is 9.5e-4 per degree C. Dividing by 1.8 to convert degrees C to degrees F, we get a coefficient of 5.2e-4 per degree F. For an increase of 5F, that's an expansion of 2.6e-3. If gasoline is $3 per gallon, the difference is 7.8e-3 dollars per gallon, that is, about 3/4 of a cent. That's an order of magnitude less than the 3 to 9 cents per gallon that people are talking about. One or the other of us has got a decimal point in the wrong place.

you have to look fairly long and hard to find an OSS project that *isn't* just a reimplementation of an existing product.

Please let us know of what commercial products the following OSS projects are mere reimplementations:

• gcc
• emacs
• TeX
• Python
• Tcl
• emdros
• redet
• Transcriber
• csound

I also use FLAC. One point to note is that the compression ratio varies considerably with the type of material compressed. I've posted comparisons of algorithms using linguistic field recordings rather than music here. For my material, the greatest compression was achieved by Lossless Audio, but the increment is not that great, and the time for both compression and decompression is much greater.

This is a variant of RadixSort, which is well known to be faster than any comparison sort such as MergeSort. The problem with non-comparison sorts is that as such they are restricted to sorting items representable a unstructured bit fields, which means, essentially, integers. A large part of the time, the real problem in sorting is (a) extracting the fields that you want to use as keys (since it is not generally the case that you want to sort on the entire record) and (b) arranging for each pair of records to compare as you need them to, which involves both recognizing the internal structure of keys (consider the case of dates) and imposing suitable weights for the individual components. In other words, in many situations the bulk of the code and time are devoted to parsing and transformation of records. So long as you are not using a really bad algorithm, the time devoted to the sort itself is likely to be a small percentage of the total time.

For example, I have written a sort utility that differs from most others in its ability to parse the input into records and records into fields and in the transformations it can apply so as to use unusual sort orders and sort on things like dates, month names, and numbers in non-Western number systems. It was originally written for alphabetizing dictionaries of "exotic" languages. It is frequently the case that the time devoted to the actual sort is less than 1% of the run time.

In sum, non-comparison sorts have a niche but are of limited utility because they get their speed from making use of additional information that is only available for a limited set of datatypes. For the great majority of applications, only comparison sorts are flexible enough to be of use.

This is true if the project is still incomplete or if it depends on other projects that are constantly changing. There are, however, projects that are complete and have had the bugs worked out. They may be updated only when someone finds a bug or something breaks because of a change in something they depend on. In such cases, no activity for months at a time is not a danger sign. I've got projects like that myself. Take my color picker. Okay, it isn't Linux or Apache, but it does what it was intended to do and has no known bugs. At some point I might add a feature or a translation, but unless something comes up, there is no need to do anything to it. The fact that the last release was nine months ago doesn't mean that it has been abandoned or that anything is wrong with it.

I agree, MS are more than justified in using the word "genuine". I don't really feel the need to jump through semantic hoops to defend the choice either. I know what they mean when they say genuine, and it's a hell of a lot easier than calling the service "Windows Copy Produced In Accordance With The Prevailing Copyright Laws In This Jurisdiction Advantage" (WCPIAWTPCLITJA for short. Trips off the tongue huh?).

On a slight side track, I really do despise these language conservatives. The meaning of words changes over time, and if enough people understand a word to mean something, then that is what it means. The compilers of the Oxford English Dictionary (I believe the equivalent reference work for you chaps in the Colonies is Websters) do not decide what a word means, they report what it means. Witness "google" becoming a verb.

On a different tangent entirely, and one somewhat more related to TFA, I think it would be instructive if people had a quick look at Bill Poser's web site. I would like in particular to direct you the list of links at the bottom "The Beginning of the Free Software Movement ", "The Free Software Foundation ", "Groklaw [Everything about SCO's anti-Linux campaign]", "LinuxLinks", "Why You Shouldn't Send People (Including Me) Microsoft Word Documents" and "Treacherous Computing". Perhaps a more fitting headline for this story would have been "Free software Advocate Finds Tenuous Excuse To Bash MS".

Just a quick kalma protection disclaimer. I use Linux every day, it's my primary desktop OS on all my machines except the Wintendo, but if Paul Thurrot wrote an article complaining about the FSF using the wrong definition of any given word, there would be 300 posts calling him an MS shill before anyone got as far as reading the article. This article is just petty MS bashing, and nothing more.

It's true that the headline is misleading. The patent application is not for "conjugating verbs". That said, it is still very broad and something for which there is tons of prior art. I don't see any specific methods mentioned. Software for generating the entire paradigm for a verb is nothing new, as various posters, including myself, have already pointed out. The idea that the user doesn't need to produce a particular citation form but that the system will figure it out is also not new. I published a paper ("Making Athabaskan Dictionaries Usable," in Gary Holton (ed.) (2002) Proceedings of the Athabaskan Languages Conference --- 2002, Fairbanks: Alaska Native Language Center, University of Alaska. Working Papers #2. pp. 136-147.) a while back about how such a system can provide a usable dictionary for languages with extremely complex verbal systems, such as Navajo.

Here's a description of an actual implementation of a system like this for Nahuatl. Unfortunately, the site at which you can actually try it out seems to be down, but it does, or at least did, exist.

You can see the disparity caused by different audiences in some sites that I run. My personal site's stats look like this:

Firefox 48.3
IE 24.2
Mozilla 9.2
Unknown 4.9
Safari 4.1
Konqueror 2.6
Opera 2.4
Netscape 1.1
Galeon 0.5
Others 1.4

Another site that I maintain is quite different:

IE 71.7
Unknown 10
Firefox 7.8
Safari 3.6
Konqueror 2.1
Mozilla 2
Netscape 1.1
Links 0.3
Opera 0.3
Others 0.6

The difference is that traffic for my personal site comes primarily from people downloading free software (most of which will run on systems other than GNU/Linux but is known mostly to Linux-oriented people), whereas the traffic for the other site, which deals with the native languages of British Columbia, presumably has no such OS or geek bias.

A third site that I manage, for some friends' fishing lodge, looks like this:

IE 74.8
Firefox 17.6
Mozilla 5.2
Unknown 2.3

If that's what you want, I don't know of anything 100% automated, but I have written something along those lines that makes the process fast. Check out: SndBite. The kind of recordings that SndBite was designed for are not as clean as language learning CDs and don't have such regular spacing between utterances, so completely automatic operation wasn't a goal. SndBite is however programmable - the init file is a Tcl program executed by a safe interpreter - so I think that this could be fully automated by exposing a few more of the internal commands in the slave interpreter.

If I had written such a poorly argued piece I wouldn't want to put my name to it, much less give my professional credentials. Take the argument about innovation. It's based on a single example! Yes, Linux is not particularly innovative. It originated as a clone, so of course it wasn't innovative. Insofar as open source attempts to replace proprietary software, there has to be a good deal of cloning. That doesn't mean that open source software is intrinsically non-innovative, just that there has been a lot of catching up to do.

Even so, software intended in the first instance to clone proprietary software has often been innovative. Many examples are to be found in the GNU project. GNU "clones" of standard Unix tools are often considered to be superior to the originals. Not only is the implementation superior (typically in having fewer bugs and fewer arbitrary limitations), but they often extend the capabilities of the original tool.

The other place in which innovation is readily seen is in areas in which there is little or no cloning activity because there is little or no proprietary software to catch up to. In my own field of linguistics, for example, there isn't a lot of proprietary software because there isn't much of a market for it. Linguists can't afford expensive software. The more interesting linguistic software that has been coming along is mostly free software. For example, the most advanced database for annotated text is emdros. It isn't a clone of anything. In phonetics the acoustic analysis program of choice currently is probably Praat. It compares favorably to commercial products. (Phonetics software is a bit different from linguistics in general in that it overlaps to a considerable extent with software for use in areas like speech pathology, where there is money to be made.) As a third example, I'll cite my own program redet, which is a regular expression search tool. It has a few features of particular interest to linguists, such as widgets for entering the International Phonetic Alphabet and the ability to intersect user-defined named character classes (which enables matching over feature matrices), but in most respects it is a regular expression tool of the same sort that programmers and various other non-linguists use. There are a number of similar free tools and at least one proprietary commercial product. However you may judge it in comparison to the others, it is unquestionably not a clone. Among its innovative features is the fact that it determines the properties of the regular expression engine that it uses empirically, by running a set of tests.

Basing a sweeping generalization on a single example is a poor practice in general, but in this case it is especially bad because Linux is an atypical example. Much open source software is innovative, and much proprietary software is not.

In fact, here are some of this papers: http://www.billposer.org/papers.html

But that's just a guess.