Net commerce works in very different ways, yet anti-trust law hasn't evolved. Microsoft didn't become a monopoly by jacking up prices, but by using practically the opposite tactic - in effect giving products away to obtain staggering market share. Gate's big idea was to make sure his company's software and operating systems were distributed so freely and aggressively they were on every desktop.
I believe you have misinterpreted what Microsoft has been doing. The only times Microsoft has given products away has been to get rid of competition. This is a classic monopoly strategy. To go back to the days of Standard Oil, Standard oil was big enough it could totally undersell the competition in one region and still make profit because it was more than balanced by its monopoly position elsewhere - where it could sell at a much higher price - and with the expectation that it would soon be the sole provider for the new region.
In the Microsoft case, this would be Microsoft's giving away of IE (in the days before it was "integrated" into the OS) to drive Netscape out of the market. Microsoft could survive because of its monopoly position in a different market, operating systems.
As another way to see this, the cost of Microsoft's software over time has increased, unlike every other component of a personal computer. Microsoft has not been giving away "freely" its two key products, Windows and Office.
Another finding of fact was that Microsoft has been using its monopoly position to tie one product to another. As you point out, you cannot get Word without Office. As Judge Jackson points out, MS used various tieing agreements to bully its own way. Again, this is another very classic monopoly power.
Neither of these monopoly tactics are affected by any fundamental change in how net commerce works.
Indeed, anti-trust almost doesn't cover most of what goes on in net commerce, because there is no (purely) net commerce company which is close to having a true monopoly position. Yes, there are companies which have a large market share, but the barriers to entry are still rather low, and none of them have the ability to exert behaviours of monopolies.
My favorite is the back door Thompson put in the unix login program. The back door wasn't in login.c but was in the C compiler, so when login was compiled, the code was added -- you never saw the code in the login source. In addition, the C compiler was modified to add the back door code when it compiled itself, so you never saw the modifications in the source.
Dionysus (thales74@yahoo.com): You're right, there are no alternative for gcc on UNIX if you want to follow the ISO C++ standard. .... How far away are we from a ISO C++ compliant compiler, anyway? Months, quarters, years?
How about KAI C++? It's been around for years. See
http://www.kai.com/C_plus_plus/index.html
From everything I've heard, it's the most ISO compliant C++ compiler you can get, and produces some of the fastest code as well. I works on many OSes, including the Linux/Intel version for $400.
Bothari (gustavo(NO)carvalho(SPAM)@(PLEASE!)mail.teleweb.p t) said: If you don't believe that fast and stable work can be done on pure C++ go check out BeOS...
The right tools for the right job, people....
BTW, read the last Be newsletter. There's an article saying, in essense, the Be kernel does not use C++. So pure C++ is not always the right tool for the job.:)
I run a website for a chemistry learning center and use Perl for just about everything.
I don't mean this as a put-down, but that's not really chemistry. How would you load one of your structures and measure, say, a dihedral angle? It is doable in Perl (I've done it) but you'll have to write your own code because no one has distributed a package for doing it.
Compare that to Tcl, where I can name several programs which let you do that, and the same for Python.
Or how about doing some molecular dynamics (MD) on it. I've seen two different Python programs which do MD, but none written in Perl.
I could list a dozen other chemical research related tasks (compute a molecular surface, identify salts in a compound, find the maximum common substructure between two chemicals, parse a SMILES string, read a mol2 file, determine connectivity, and more). I've never seen a Perl program available which can do these, even with extensions written in C/C++. I have seen various of these in both Tcl and Python.
It isn't that Perl cannot be used for these tasks, but I believe strongly that Perl is not very suitable to chemistry work. Although negative proofs are tricky, I use as evidence the relative paucity of Perl package for doing chemistry, compared to other languages.
There are more chemistry programs which implement their own language from scratch than use Perl!
Now, Perl is often used to drive these programs in a rather loose fashion, and glue them with other systems, as you have done with your site, but this is not the same as doing chemistry with Perl.
P.S. I didn't know that Chime could do atomic orbitals; RasMol, from which it is based, cannot, and I don't see anything at mdli.com or the umass site describing the feature or even suggesting it. Do you mean VdW spheres, or if not, could you point me to a description of it? Thanks!
LizardKing said: Most of the code behind the unravelling of DNA is written in Perl. And they even make it freelly available...
Sure, that's the bioperl effort, as well as some related projects. You can even see some of my contributions at bioperl.org. However, that's bioinformatics and not chemistry.
I've yet to see Perl used intensively for building chemical compounds, or doing molecular dynamics simulations, or doing substructure searches, or visualizing 3D structures. I have seen Python and Tcl code for those tasks.
To repeat myself, Perl is used a lot in bioinformatics, but rarely in chemistry.
Interestingly, there are very few people using Perl for chemistry. I know of DayPerl, used at a couple of sites, but there is also PyDaylight (which I helped write). I can't think of *any* other examples. There are some proposals in bioperl.org, but no code for chemistry.
Most of the programs either have their own language or embed Tcl. There is a RasMol variant - which Larry Wall used to make the images -- which uses Tcl. Cerius2 (a commercial package) uses Tcl. CACTVS uses Tcl for small molecule chemistry, as does VMD for molecular visualization (which I also helped write). MMTK uses Python for biomolecular modeling. I've even seen Prolog used for one system!
But never Perl. Perl is used a lot in bioinformatics, but rarely in chemistry.
If you have pointers, please contact me at dalke@acm.org. One of my interests is in how people use very-high level languages for computational chemistry and biology.
theres always a difference between 2 frames. they arent exactly alike due to variances in light...
Story time.
About 8 years ago I was doing some lab work with a frame grapper. We used it to find if, at a given voltage, the sample had undergone a phase change or not. As you approach the critical voltage, the time to phase change takes longer and longer.
So we automated it by comparing frames over time with the intial frame. Problem was, the grabber would get one of the two interleaves, so we had to take three snapshots and use the smallest of the differences |2-1|, |3-1| and |3-2|.
In other words, even if the scene doesn't change you could still get some differences between the images.
3. Check into RCS. 4. Compile code. 5. If necessary, debug with gdb.
You put untested (and on average, likely broken) code into your code repository? You must not be working with many other people. Else you would have people wanting your head on a platter because you busted their compilation after they did an update, making them waste time tracking down errors in your code.
These stories are only apocryphal, having heard them some 12 years ago. They are meant to show that these sorts of studies are hard to interpret.
One of the stories concerns mouse longevity. In the study, mice were irradiated to see if they would live longer or shorter than the control mice. Turns out they lived longer, so the first conclusion was that radiation could be beneficial.
After more work, they concluded that the mice they were using had a high rate of mortality when giving birth. What the study was seeing was that the irradiated mice were more often sterile, and hence living longer.
The second story was something like observing that mice drank water less while exposed to radiation. In this case, what happened was the water bottle used in the radiation cages was a different type of plastic (more resistent to radiation) than the normal cages. The mice didn't like the taste of the plastic, so didn't drink as often.
Moral: be wary of mice which glow in the dark:) (and yes, I know radioactive mice don't glow).
For example, one of them latched onto the fact that CVS doesn't have exclusive checkouts like VSS and he claims it will make it impossible to work as a team if we can't know exactly who has a file checked out at any given time.
As I recall, the CVS documentation says that if this is ever a problem then there is a problem in your process. People should know what others are working on. The doc go into more detail. that.
Speaking for myself, I've worked with RCS which has pessimistic file locking like you mentioned. I found it more of a hassle finding a file was locked and having to bug people to check in their code, or breaking the lock because that person was on vacation and forgot to check everything in.
On the contrary, a good habit to be in is to make small changes to the source and check it in. That way you discover conflicts quickly. It also requires you have good modularity in your projects, so changes are usually limited to one file and not scattered everywhere. For those cases which do require many large changes, they are big enough that you would have to tell everyone to stop working anyway (ie, under VSS you have to tell everyone to check their code back in.)
Also, I have some concerns about how the automatic three-way merging will work with binary form files (can't do a diff/merge then) which our Oracle work uses.
Take a look at the documentation section on binary files. There's an option, `-kb', to specify that a given file is a binary file, and there are options in cvswrappers(?) to specify how to to the merging. I believe you can specify your own way to do that.
Re:Wow! No longer need CLUELESS Journalists
on
Buffy and Dr. Varnus
·
· Score: 1
As I can't afford the scientific journals I rely on the popular press.
Try your local public library. Most subscribe to either Science or Nature.
dillon_rinker said: If I were really mean, once I'd done all my R&D for my original wipers, I would start the patent process, and drag it out r-e-a-l s-l-o-w..... This kind of patent is known as a submarine patent, and many patent lawyers have gotten rich this way. Submarine patents are BAD.
As I recall, the latest change to the patent law says that the patent is valid for 20 years after date of submission, as compared to 17 years after receiving the patent. This was partially to close the submarine patent loophole.
k_wayne@linuxpower.org said At the end of the day, 99% of the best code in the world is going to be GPL'ed, as anything less is a lame attempt to keep some form of control on the code.
Just pointing out that "not closed source" doesn't always mean "GPL." Python is distributed with the Python license, which is of the "use, copy, modify, and distribute" type. The README specifically says:
The Python distribution is *not* affected by the GNU Public Licence (GPL). There are interfaces to some GNU code but these are entirely optional and no GNU code is distributed with Python.
if he went into the time thingamabob 1000 years on Jan 1, 2000 how could he have come out on Dec 31, 2999?
If you're going to be picky, you could also ask why it was daytime instead of being a few seconds from midnight. Here's a couple reasons: The earth's rotation is slowly decreasing (hence some 7 leap seconds over the last couple decades). The timer in the cryofreezer wasn't exactly accurate for a span of 1000 years (that's off by 1 in 2*10-06!).
If you want to be more picky, the first time old New York was rebuilt, you see some buildings (castles?). Why weren't they underground instead of the original New York?
Anonymous Shepherd (louisjr@cco.caltech.edu) said: the S-shaped growth curve does exist, in a way, because what Rob & Co do is review submissions and then post them, much like friends forwarding links to friends, and in this way, Rob & Co forward links to their friends, us.
To be a good S shaped curve, you and I would have to tell roughly as many friends as Slashdot has readers. I know I don't! So the effect is strongly driven by Rob & Co. and only weakly affected by us. But I don't know how weak "weakly" is.
other news sites and such who frequent/. may also post interesting links
If you take a look at the graphs I referenced you'll see some consequence of that effect. Some of the download freqencies overlap. Still, to see real evidence of that you'll likely need to watch some place with a much broader audience, like CNN.com when the US started the attacks.
Otherwise I suspect that the peak timescale (which seems to be about an hour or two) is smaller than the combination of 1) the time it takes for an article to be rewritten/posted/linked at another news site and 2) the number of different sites people visit in a day. (I can only stay on top of about 4 sites.)
Are you two both bi-chem majors or something?
I guess "something." I'm no longer. a student. If you really want to know what I do, you've enough info to web search me and ask direct.
NumberCruncher (landman@mediaone.net) said: I dont think anyone here cares much about floppy modes of macromolecules, or CCSD calculations, or non-uniform sampling of distributions in Monte Carlo simulations in order to examine rare events (some of the talks I went to).
You're right. I'm much more interesting in conformational modes of floppy molecules (as for pharmacophore prediction). And CCSD? I try to stay away from ab initio work and stick with classical molecular dynamics. Regarding the last of your three, we leaned more towards steered umbrella sampling, and the closest thing to Monte Carlo was the Langevin term.
:)
think of the slashdot effect (SE) as an avalanche, or a statistical mechanical critical exponent problem.
I model it more like queuing theory. There is a population of N people checking Slashdot. Different people check/. at different time intervals, which creates a distribution of checking frequencies n(t), most likely with a Poisson distribution.
I believe most people will follow a link the first time it comes up, though this also has a distribution. (A few people will check a link several times, but the effect is lessened somewhat by caching.)
Thus, I would expect to see a roughly t*exp(-t) shape to the Slashdot effect, so it should start linear and have an exponential tail.
As it turns out, there is some data to test this theory. Alas, it isn't very good data given the large-ish bin size and the existance of the data in graph form only. It looks like it can be eyeball fit by a Poisson function.
Still, I don't really see an exponential growth curve as you suggest (though again it's hard to tell) and I don't intuitively feel that your description is correct.
Ahh, there is a possibility. If I view a link then forward to friends, who forward it to friends, etc, then there should be an S-shaped growth curve. But I only see about a couple percent of the/. articles forward to the different groups and mailing lists I read. (Though extrapolation from one datum is rather imprecise:).
Still, I assert it's less precise to describe it as a critical exponent problem than as a Poisson distribution.
The femto-laser fissions Uranium, it doesn't fuse anything... Both/. and the news site have it wrong, unless they're leaving something out of the article.
I do believe you only read the box at the end of the article. The story is on D+D->He fusion.
But speaking of which, I was at an Open House in the Univ. of Illinois (UIUC) physics department in 1994. They were showing off a small fusion reactor (a sphere about a meter across) which accelerates plasma into the center, and causes fusion and hence produces neutrons. The idea was to use this source to calibrate the detectors for Princeton's (?) tokamak reactor.
In other words, small scale fusion reactors already exist, so why is this one special, other than being novel for using a laser for acceleration instead of an high-voltage electric field?
an AC said Why use a language that you hate? I am a big Python fan, having switched from several years of Perl programming. I don't hate Perl, so I cannot give a real answer, but I will say that sometimes:
#!/usr/local/bin/perl -pw BEGIN { %color = ( "A" => "red", "T" => "yellow", "C" => "blue", "G" => "green"); } s|([ATCG])|"<font color='$color{$1}'>$1</font>"|eg if !/^>/;
is just too cool for words (this HTML colorizes input DNA sequences in FASTA format based on the residue type).
Doesn't Python have a regular expression library? Sure does, and Perl5 compatible (excepting a few examples with backtracking based on embedded perl code). However, the equivalent would be somewhat more complicated.
Also in Dylan, I'm sure this re example would be more complicated in Dylan as well.
methods can return more than one value. Perl and Python both allow multiple return values, so I don't understand the relevance of your comment.
The graph given in the report showing the exponential growth in the kernel code is at odds with the data from: http://ps.cus.umist.ac.uk/~rhw/kernel.versions.htm l . Using the tar.gz files listed therein (granted, it is compressed) I made a graph of the kernels over time. Here's the results in ASCII (my apologies for the size and formatting): Size in MB 12 ++----+-----------+----------+----------+--------- --+----++ + + + + + + + | A | 10 ++ AA ++ | AAA | | AAA | | AAA | 8 ++ AA ++ | AA | | AAA | 6 ++ AAA ++ | AA AA A | | AAAAAA AAA | | AAA | 4 ++ A ++ | AAA | | AAA | 2 ++ AAA ++ | AAAAAAA | | AAAAAAA | + + AAAAAAAAA + + + + 0 0 ++-AAAAAAAA-------+----------+----------+--------- --+----++
1992 1993 1994 1995 1996 1997 1998 1999 There is no exponential growth in the archive file sizes. I've got graphics graphs, but no site to host them. Anyone want to contribute web space for my extraction source code and graphics?
The author has obvious biases all throught the essay which make it feel more like a revisionist essay than an attempt at history. Here are some examples:
To accomplish this, they turned to a more than fifty year old fad, Why the word "fad" over "practice?" The former is often used to indicate something short lived and with little long term effect.
In the computer field, this spelled, among other things, the introduction of programming standards, code reviews, structured walkthroughs and miscellaneous programming productivity metrics. Interesting. I've found code reviews and walkthroughs very useful as a way to get peoples' code to work together. So this is a Taylorist fad? I don't plan to stop, esp. since every review I've been in has found bugs in the code. However, metrics are not very useful.
It also prevented programmers from learning very much about the workings of the machine they programmed. We also use code reviews as a socialization method so others can learn about how things work.
The emergence of hackers as an identifiable group coincides closely in time with the introduction of various Taylorist methods in software development. He has established correlation, not causation. You have to have many people to be a recognizable group, and these people have to work somewhere.
Structured Programming is often the buzzword for an attempt to routinize and deskill programming work I believe the nomenclature has changed over the decades. I will not work with someone who does not write structured code.
The hacker had been forced by his programming-illiterate boss into using some tools and methods that he considered unsuitable or inadequate for the task at hand. Or believes programming is a strictly personal and artistic expression and doesn't want to learn how to write code usable by others.
The Hacker Community... A previous respondent said this was a full, detailed history of hackerdom. It isn't. At the very least, SAIL (the Stanford AI Lab) is not mentioned.
The next step in establishing a hacking community was the ARPANet The author really must include dates in his essay. I seem to recall ARPA starting in the early 1970s. Does this mean the hackers of the 60s and 50s (both of which he asserts existed) did not comprise a community? They were.
Consequently, first wave hackers became some of the most vigorous members in the communities commissioned to develop the ARPANet. Again, dates would be nice. Who are the first wave hackers?
In contrast, the hacker's fascination with technology is not because they believe that technology will bring about great and revolutionary changes (or make any societal difference whatsoever) You must be aware of the possible effects of technology on society. Not doing so is a sign of irresponsibility. That's why organizations like the ACM have a code of ethics.
Hackers love technology for its own sake. But hackers believe that technology is too a good a thing to be proprietary. I find these two viewpoints somewhat contradictory. What should a hacker do when offered to work with some really cool patented technology which will remain proprietary? If the love of technology is all, then there's no quandary -- work with the technology. This is not the case.
While the techno-yuppies usually portrait the Internet as a laissez-faire utopia created out of thin air, If I am to take the reference to Nicholas Negroponte as an example of a "techno-yuppy" then this statment is wrong. Looking away from computers, you will see the same commentary about nuclear power in the 50s -- "too cheap to meter" -- when people who knew about nuclear power knew better. This "techno-yuppy" viewpoint comes from incomplete understanding, not some fundamental belief.
Neither do most hackers share the techno-yuppies view of the Internet as a lawless environment Ha! The idea of "lawless environment" was the common belief amoung hackers some five years ago. In general, most people starting with the Internet believe there is some agency with regulartory power. The author must provide some way to back up his claim, for I believe it to be false.
If necessary, hackers may at times do consulting "for money" just to make ends meet, but their priorities are clearly to earn a living in order to create beautiful technology, not the other way around. My. That's an old-style view of the world. I consider myself a hacker. I work in a company. I like working in a company because, believe it or not, I have more freedom to do things than in a university. And, I get to work on "free" software. Or, ask Alan Cox or any of the other non-academic hackers in the world. Now, can I have some numbers to back up his claim?
Within the Hacker Community Is POSIX within or without? Most people I know perfer XML as an exchange document over HTML and don't think of it as a multimedia format. LaTeX is "outside the hacker community?" Could have fooled me.
It is, however, interesting to note that the two most dysfunctional and misengineered...originated from semi-democratic processes And what of ANSI C? And C++? And POSIX? And HTML 3 (or whatever number we're at)? And X11 is also one quite screwed up library (say, compared to DisplayPostscript), but that's within the "hacker community." I've tried understanding how gcc works, but quickly became confused.
For typesetting complex text and graphics, the hacker community has provided TeX. The hacker community? As I recall, Knuth did most of TeX in the late 1970s, and he started using computers in the 1950s. Thus, are his ethics "first wave" or some other term?
One of the reasons it does not behave like that by default is because the text editor is one of the programs that receives the most intensive use in many computing environments Now that is a severe case of revisionism. I have never heard anyone use that arguement before. Can someone point this out to me in a FAQ or emacs document somewhere? Oh, and I use xemacs, which is a GUI as WordPad or nedit.
Commercial software is generally produced by teams of computer workers churning out the various components to a pre-set list of specifications. Again, the author's bias is evidenced by using the word "churning." If he had read a good software engineering book (and "Rapid Development" is my favorite) he would easily have found that some types of projects are best done that way; those for which the end results are well known. I imagine if there is ever a gcc rewrite, it will be done this way.
it is unlikely that the computer workers themselves or anyone they actually know will make extensive use of what he creates. This statement has little meaning. Does this mean that someone writing software for rocket avionics must be a rocket scientist? Or that someone working on pacemaker software must need a pacemaker?
As a result, the functional properties and qualities of the finished artifact are of little concern to the worker. Sure, but metal workers these days are not managed by a Tayloristic scheme and metal workers do take pride in thier work. My own grandfather did metal machining for a living and he was proud of what he did. So it is possible for people to do good work on things for which they are not the end user.
Sometimes the myopia this causes lead to serious errors of judgement Hold up your hands people who haven't had serious errors of judgement for your own code? Thought so.
A hacker, on the other hand, does not perform well producing piecework software components based upon pre-set specifications. So the authors of glibc are not hackers? And the people working on NFS3 support Linux? And what about those who write drivers for new file systems, like NTFS? In all those cases there are pre-set specifications.
hackers generally require access to the artifact's interior and want to understand all parts of the systems they work on, including the components contributed by others. This is just as true for "workers in programming teams." Or, is a prerequiste of hackers that they cannot work in teams?
the difference between the hacker's approach and those of the industrial programmer is one of outlook: between an agoric, integrated and holistic attitude towards the creation of artifacts and a proprietary, fragmented and reductionist one. I work in industry. I want to see code. He has asserted a false dichotomy.
Hacking as a method for system development originated as a grass-roots reaction to attempts to impose an industrial mode of production on the development of software. Here the author confuses the word "Taylorism" (which is now rarely practiced in its original form) with "industrial." Couple of questions: Are the Mozilla developers, who apply pretty standard industry practices (proposals, designs, bug tracking, build management, etc.) hackers? Are software companies which have non-existent internal management full of hackers?
Here's an expansion of the comments I made on intellectualcapital.com:
Eric Raymond says there are two styles of software distribution, "closed source" and "open source." He equivalences the latter with "the Internet engineering tradition" and says they are "fundamentally opposed styles of software development." I believe he is wrong. The world is not polarized into two camps; there are many intermediate forms.
Consider a commerical product where customers get the software and are allowed to modify it but not redistribute the source or modifications. This can be loosened someone by allowing patches to be redistributed gratis to other licensed customers. In this case, the customer gets nearly all of the benefits of "open source" (peer review, a fallback path if the supplier stops working on the project, etc.) without the need for granting redistribution rights.
I worked on one proprietary software package which was released with source code under a license similar to that I gave, so this does happen in real life.
I see very, very few additional advantages to the *customer* between this model and the "open source" model advocated by Raymond. I have never heard him address the need for nor benefit of the redistribution requirement.
On the argument he gave, that of a company which just spent several million on closed, propritary software. He forgot to mention that the normal solution to that problem is for the supplier to put the source code in escrow. If the supplier goes out of business or can no longer provide support or revisions, the customer gets the source code. This is standard.
Finally, he omits to mention that some "closed source" software has been found to be of very high quality. There's a very famous study of the quality of the space shuttle avionics code. As I recall, there was one problem found in the integration level, and it wasn't serious enough to affect the mission. The problem is this sort of code review is very, very expensive. I believe it was in the many hundreds of dollars per line of code.
Slashdot Mirror
I believe you have misinterpreted what Microsoft has been doing. The only times Microsoft has given products away has been to get rid of competition. This is a classic monopoly strategy. To go back to the days of Standard Oil, Standard oil was big enough it could totally undersell the competition in one region and still make profit because it was more than balanced by its monopoly position elsewhere - where it could sell at a much higher price - and with the expectation that it would soon be the sole provider for the new region.
In the Microsoft case, this would be Microsoft's giving away of IE (in the days before it was "integrated" into the OS) to drive Netscape out of the market. Microsoft could survive because of its monopoly position in a different market, operating systems.
As another way to see this, the cost of Microsoft's software over time has increased, unlike every other component of a personal computer. Microsoft has not been giving away "freely" its two key products, Windows and Office.
Another finding of fact was that Microsoft has been using its monopoly position to tie one product to another. As you point out, you cannot get Word without Office. As Judge Jackson points out, MS used various tieing agreements to bully its own way. Again, this is another very classic monopoly power.
Neither of these monopoly tactics are affected by any fundamental change in how net commerce works.
Indeed, anti-trust almost doesn't cover most of what goes on in net commerce, because there is no (purely) net commerce company which is close to having a true monopoly position. Yes, there are companies which have a large market share, but the barriers to entry are still rather low, and none of them have the ability to exert behaviours of monopolies.
See http://www.acm.org/classics/sep95/for more details.
You're right, there are no alternative for gcc on UNIX if you want to follow the ISO C++ standard.
How far away are we from a ISO C++ compliant compiler, anyway? Months, quarters, years?
How about KAI C++? It's been around for years. See
http://www.kai.com/C_plus_plus/index.html
From everything I've heard, it's the most ISO compliant C++ compiler you can get, and produces some of the fastest code as well. I works on many OSes, including the Linux/Intel version for $400.
If you don't believe that fast and stable work can be done on pure C++ go check out BeOS...
The right tools for the right job, people....
BTW, read the last Be newsletter. There's an article saying, in essense, the Be kernel does not use C++. So pure C++ is not always the right tool for the job. :)
You might be surprised.
I run a website for a chemistry learning center and use Perl for just about everything.
I don't mean this as a put-down, but that's not really chemistry. How would you load one of your structures and measure, say, a dihedral angle? It is doable in Perl (I've done it) but you'll have to write your own code because no one has distributed a package for doing it.
Compare that to Tcl, where I can name several programs which let you do that, and the same for Python.
Or how about doing some molecular dynamics (MD) on it. I've seen two different Python programs which do MD, but none written in Perl.
I could list a dozen other chemical research related tasks (compute a molecular surface, identify salts in a compound, find the maximum common substructure between two chemicals, parse a SMILES string, read a mol2 file, determine connectivity, and more). I've never seen a Perl program available which can do these, even with extensions written in C/C++. I have seen various of these in both Tcl and Python.
It isn't that Perl cannot be used for these tasks, but I believe strongly that Perl is not very suitable to chemistry work. Although negative proofs are tricky, I use as evidence the relative paucity of Perl package for doing chemistry, compared to other languages.
There are more chemistry programs which implement their own language from scratch than use Perl!
Now, Perl is often used to drive these programs in a rather loose fashion, and glue them with other systems, as you have done with your site, but this is not the same as doing chemistry with Perl.
P.S. I didn't know that Chime could do atomic orbitals; RasMol, from which it is based, cannot, and I don't see anything at mdli.com or the umass site describing the feature or even suggesting it. Do you mean VdW spheres, or if not, could you point me to a description of it? Thanks!
Most of the code behind the unravelling of DNA is written in Perl. And they even make it freelly available
Sure, that's the bioperl effort, as well as some related projects. You can even see some of my contributions at bioperl.org. However, that's bioinformatics and not chemistry.
I've yet to see Perl used intensively for building chemical compounds, or doing molecular dynamics simulations, or doing substructure searches, or visualizing 3D structures. I have seen Python and Tcl code for those tasks.
To repeat myself, Perl is used a lot in bioinformatics, but rarely in chemistry.
Interestingly, there are very few people using Perl for chemistry. I know of DayPerl, used at a couple of sites, but there is also PyDaylight (which I helped write). I can't think of *any* other examples. There are some proposals in bioperl.org, but no code for chemistry.
Most of the programs either have their own language or embed Tcl. There is a RasMol variant - which Larry Wall used to make the images -- which uses Tcl. Cerius2 (a commercial package) uses Tcl. CACTVS uses Tcl for small molecule chemistry, as does VMD for molecular visualization (which I also helped write). MMTK uses Python for biomolecular modeling. I've even seen Prolog used for one system!
But never Perl. Perl is used a lot in bioinformatics, but rarely in chemistry.
If you have pointers, please contact me at dalke@acm.org. One of my interests is in how people use very-high level languages for computational chemistry and biology.
Story time.
About 8 years ago I was doing some lab work with a frame grapper. We used it to find if, at a given voltage, the sample had undergone a phase change or not. As you approach the critical voltage, the time to phase change takes longer and longer.
So we automated it by comparing frames over time with the intial frame. Problem was, the grabber would get one of the two interleaves, so we had to take three snapshots and use the smallest of the differences |2-1|, |3-1| and |3-2|.
In other words, even if the scene doesn't change you could still get some differences between the images.
4. Compile code.
5. If necessary, debug with gdb.
You put untested (and on average, likely broken) code into your code repository? You must not be working with many other people. Else you would have people wanting your head on a platter because you busted their compilation after they did an update, making them waste time tracking down errors in your code.
One of the stories concerns mouse longevity. In the study, mice were irradiated to see if they would live longer or shorter than the control mice. Turns out they lived longer, so the first conclusion was that radiation could be beneficial.
After more work, they concluded that the mice they were using had a high rate of mortality when giving birth. What the study was seeing was that the irradiated mice were more often sterile, and hence living longer.
The second story was something like observing that mice drank water less while exposed to radiation. In this case, what happened was the water bottle used in the radiation cages was a different type of plastic (more resistent to radiation) than the normal cages. The mice didn't like the taste of the plastic, so didn't drink as often.
Moral: be wary of mice which glow in the dark :) (and yes, I know radioactive mice don't glow).
As I recall, the CVS documentation says that if this is ever a problem then there is a problem in your process. People should know what others are working on. The doc go into more detail. that.
Speaking for myself, I've worked with RCS which has pessimistic file locking like you mentioned. I found it more of a hassle finding a file was locked and having to bug people to check in their code, or breaking the lock because that person was on vacation and forgot to check everything in.
On the contrary, a good habit to be in is to make small changes to the source and check it in. That way you discover conflicts quickly. It also requires you have good modularity in your projects, so changes are usually limited to one file and not scattered everywhere. For those cases which do require many large changes, they are big enough that you would have to tell everyone to stop working anyway (ie, under VSS you have to tell everyone to check their code back in.)
Also, I have some concerns about how the automatic three-way merging will work with binary form files (can't do a diff/merge then) which our Oracle work uses.
Take a look at the documentation section on binary files. There's an option, `-kb', to specify that a given file is a binary file, and there are options in cvswrappers(?) to specify how to to the merging. I believe you can specify your own way to do that.
Try your local public library. Most subscribe to either Science or Nature.
If I were really mean, once I'd done all my R&D for my original wipers, I would start the patent process, and drag it out r-e-a-l s-l-o-w.
As I recall, the latest change to the patent law says that the patent is valid for 20 years after date of submission, as compared to 17 years after receiving the patent. This was partially to close the submarine patent loophole.
At the end of the day, 99% of the best code in the world is going to be GPL'ed, as anything less is a lame attempt to keep some form of control on the code.
Just pointing out that "not closed source" doesn't always mean "GPL." Python is distributed with the Python license, which is of the "use, copy, modify, and distribute" type. The README specifically says:
The Python distribution is *not* affected by the GNU Public Licence (GPL). There are interfaces to some GNU code but these are entirely optional and no GNU code is distributed with Python.
If you're going to be picky, you could also ask why it was daytime instead of being a few seconds from midnight. Here's a couple reasons: The earth's rotation is slowly decreasing (hence some 7 leap seconds over the last couple decades). The timer in the cryofreezer wasn't exactly accurate for a span of 1000 years (that's off by 1 in 2*10-06!).
If you want to be more picky, the first time old New York was rebuilt, you see some buildings (castles?). Why weren't they underground instead of the original New York?
the S-shaped growth curve does exist, in a way, because what Rob & Co do is review submissions and then post them, much like friends forwarding links to friends, and in this way, Rob & Co forward links to their friends, us.
To be a good S shaped curve, you and I would have to tell roughly as many friends as Slashdot has readers. I know I don't! So the effect is strongly driven by Rob & Co. and only weakly affected by us. But I don't know how weak "weakly" is.
other news sites and such who frequent /. may also post interesting links
If you take a look at the graphs I referenced you'll see some consequence of that effect. Some of the download freqencies overlap. Still, to see real evidence of that you'll likely need to watch some place with a much broader audience, like CNN.com when the US started the attacks.
Otherwise I suspect that the peak timescale (which seems to be about an hour or two) is smaller than the combination of 1) the time it takes for an article to be rewritten/posted/linked at another news site and 2) the number of different sites people visit in a day. (I can only stay on top of about 4 sites.)
Are you two both bi-chem majors or something?
I guess "something." I'm no longer. a student. If you really want to know what I do, you've enough info to web search me and ask direct.
I dont think anyone here cares much about floppy modes of macromolecules, or CCSD calculations, or non-uniform sampling of distributions in Monte Carlo simulations in order to examine rare events (some of the talks I went to).
You're right. I'm much more interesting in conformational modes of floppy molecules (as for pharmacophore prediction). And CCSD? I try to stay away from ab initio work and stick with classical molecular dynamics. Regarding the last of your three, we leaned more towards steered umbrella sampling, and the closest thing to Monte Carlo was the Langevin term.
think of the slashdot effect (SE) as an avalanche, or a statistical mechanical critical exponent problem.
I model it more like queuing theory. There is a population of N people checking Slashdot. Different people check /. at different time intervals, which creates a distribution of checking frequencies n(t), most likely with a Poisson distribution.
I believe most people will follow a link the first time it comes up, though this also has a distribution. (A few people will check a link several times, but the effect is lessened somewhat by caching.)
Thus, I would expect to see a roughly t*exp(-t) shape to the Slashdot effect, so it should start linear and have an exponential tail.
As it turns out, there is some data to test this theory. Alas, it isn't very good data given the large-ish bin size and the existance of the data in graph form only. It looks like it can be eyeball fit by a Poisson function.
Still, I don't really see an exponential growth curve as you suggest (though again it's hard to tell) and I don't intuitively feel that your description is correct.
Ahh, there is a possibility. If I view a link then forward to friends, who forward it to friends, etc, then there should be an S-shaped growth curve. But I only see about a couple percent of the /. articles forward to the different groups and mailing lists I read. (Though extrapolation from one datum is rather imprecise :).
Still, I assert it's less precise to describe it as a critical exponent problem than as a Poisson distribution.
Both
I do believe you only read the box at the end of the article. The story is on D+D->He fusion.
But speaking of which, I was at an Open House in the Univ. of Illinois (UIUC) physics department in 1994. They were showing off a small fusion reactor (a sphere about a meter across) which accelerates plasma into the center, and causes fusion and hence produces neutrons. The idea was to use this source to calibrate the detectors for Princeton's (?) tokamak reactor.
In other words, small scale fusion reactors already exist, so why is this one special, other than being novel for using a laser for acceleration instead of an high-voltage electric field?
I must admit, I found the pre-Star Wars-release articles quite funny when they tried to describe "Force", "Jedi Knight" and "wookie."
Why use a language that you hate?
I am a big Python fan, having switched from several years of Perl programming. I don't hate Perl, so I cannot give a real answer, but I will say that sometimes:
#!/usr/local/bin/perl -pw
BEGIN {
%color = ( "A" => "red", "T" => "yellow", "C" => "blue", "G" => "green");
}
s|([ATCG])|"<font color='$color{$1}'>$1</font>"|eg if !/^>/;
is just too cool for words (this HTML colorizes input DNA sequences in FASTA format based on the residue type).
Doesn't Python have a regular expression library?
Sure does, and Perl5 compatible (excepting a few examples with backtracking based on embedded perl code). However, the equivalent would be somewhat more complicated.
Also in Dylan,
I'm sure this re example would be more complicated in Dylan as well.
methods can return more than one value.
Perl and Python both allow multiple return values, so I don't understand the relevance of your comment.
The graph given in the report showing the exponential growth in the kernel code is at odds with the data from: http://ps.cus.umist.ac.uk/~rhw/kernel.versions.htm l . Using the tar.gz files listed therein (granted, it is compressed) I made a graph of the kernels over time. Here's the results in ASCII (my apologies for the size and formatting): Size in MB- --+----++- --+----++
12 ++----+-----------+----------+----------+--------
+ + + + + + +
| A |
10 ++ AA ++
| AAA |
| AAA |
| AAA |
8 ++ AA ++
| AA |
| AAA |
6 ++ AAA ++
| AA AA A |
| AAAAAA AAA |
| AAA |
4 ++ A ++
| AAA |
| AAA |
2 ++ AAA ++
| AAAAAAA |
| AAAAAAA |
+ + AAAAAAAAA + + + +
0 0 ++-AAAAAAAA-------+----------+----------+--------
1992 1993 1994 1995 1996 1997 1998 1999
There is no exponential growth in the archive file sizes. I've got graphics graphs, but no site to host them. Anyone want to contribute web space for my extraction source code and graphics?
To accomplish this, they turned to a more than fifty year old fad,
Why the word "fad" over "practice?" The former is often used to indicate something short lived and with little long term effect.
In the computer field, this spelled, among other things, the introduction of programming standards, code reviews, structured walkthroughs and miscellaneous programming productivity metrics.
Interesting. I've found code reviews and walkthroughs very useful as a way to get peoples' code to work together. So this is a Taylorist fad? I don't plan to stop, esp. since every review I've been in has found bugs in the code. However, metrics are not very useful.
It also prevented programmers from learning very much about the workings of the machine they programmed.
We also use code reviews as a socialization method so others can learn about how things work.
The emergence of hackers as an identifiable group coincides closely in time with the introduction of various Taylorist methods in software development.
He has established correlation, not causation. You have to have many people to be a recognizable group, and these people have to work somewhere.
Structured Programming is often the buzzword for an attempt to routinize and deskill programming work
I believe the nomenclature has changed over the decades. I will not work with someone who does not write structured code.
The hacker had been forced by his programming-illiterate boss into using some tools and methods that he considered unsuitable or inadequate for the task at hand.
Or believes programming is a strictly personal and artistic expression and doesn't want to learn how to write code usable by others.
The Hacker Community...
A previous respondent said this was a full, detailed history of hackerdom. It isn't. At the very least, SAIL (the Stanford AI Lab) is not mentioned.
The next step in establishing a hacking community was the ARPANet
The author really must include dates in his essay. I seem to recall ARPA starting in the early 1970s. Does this mean the hackers of the 60s and 50s (both of which he asserts existed) did not comprise a community? They were.
Consequently, first wave hackers became some of the most vigorous members in the communities commissioned to develop the ARPANet.
Again, dates would be nice. Who are the first wave hackers?
In contrast, the hacker's fascination with technology is not because they believe that technology will bring about great and revolutionary changes (or make any societal difference whatsoever)
You must be aware of the possible effects of technology on society. Not doing so is a sign of irresponsibility. That's why organizations like the ACM have a code of ethics.
Hackers love technology for its own sake. But hackers believe that technology is too a good a thing to be proprietary.
I find these two viewpoints somewhat contradictory. What should a hacker do when offered to work with some really cool patented technology which will remain proprietary? If the love of technology is all, then there's no quandary -- work with the technology. This is not the case.
While the techno-yuppies usually portrait the Internet as a laissez-faire utopia created out of thin air,
If I am to take the reference to Nicholas Negroponte as an example of a "techno-yuppy" then this statment is wrong. Looking away from computers, you will see the same commentary about nuclear power in the 50s -- "too cheap to meter" -- when people who knew about nuclear power knew better. This "techno-yuppy" viewpoint comes from incomplete understanding, not some fundamental belief.
Neither do most hackers share the techno-yuppies view of the Internet as a lawless environment
Ha! The idea of "lawless environment" was the common belief amoung hackers some five years ago. In general, most people starting with the Internet believe there is some agency with regulartory power. The author must provide some way to back up his claim, for I believe it to be false.
If necessary, hackers may at times do consulting "for money" just to make ends meet, but their priorities are clearly to earn a living in order to create beautiful technology, not the other way around.
My. That's an old-style view of the world. I consider myself a hacker. I work in a company. I like working in a company because, believe it or not, I have more freedom to do things than in a university. And, I get to work on "free" software. Or, ask Alan Cox or any of the other non-academic hackers in the world. Now, can I have some numbers to back up his claim?
Within the Hacker Community
Is POSIX within or without? Most people I know perfer XML as an exchange document over HTML and don't think of it as a multimedia format. LaTeX is "outside the hacker community?" Could have fooled me.
It is, however, interesting to note that the two most dysfunctional and misengineered ...originated from semi-democratic processes
And what of ANSI C? And C++? And POSIX? And HTML 3 (or whatever number we're at)? And X11 is also one quite screwed up library (say, compared to DisplayPostscript), but that's within the "hacker community." I've tried understanding how gcc works, but quickly became confused.
For typesetting complex text and graphics, the hacker community has provided TeX.
The hacker community? As I recall, Knuth did most of TeX in the late 1970s, and he started using computers in the 1950s. Thus, are his ethics "first wave" or some other term?
One of the reasons it does not behave like that by default is because the text editor is one of the programs that receives the most intensive use in many computing environments
Now that is a severe case of revisionism. I have never heard anyone use that arguement before. Can someone point this out to me in a FAQ or emacs document somewhere? Oh, and I use xemacs, which is a GUI as WordPad or nedit.
Commercial software is generally produced by teams of computer workers churning out the various components to a pre-set list of specifications.
Again, the author's bias is evidenced by using the word "churning." If he had read a good software engineering book (and "Rapid Development" is my favorite) he would easily have found that some types of projects are best done that way; those for which the end results are well known. I imagine if there is ever a gcc rewrite, it will be done this way.
it is unlikely that the computer workers themselves or anyone they actually know will make extensive use of what he creates.
This statement has little meaning. Does this mean that someone writing software for rocket avionics must be a rocket scientist? Or that someone working on pacemaker software must need a pacemaker?
As a result, the functional properties and qualities of the finished artifact are of little concern to the worker.
Sure, but metal workers these days are not managed by a Tayloristic scheme and metal workers do take pride in thier work. My own grandfather did metal machining for a living and he was proud of what he did. So it is possible for people to do good work on things for which they are not the end user.
Sometimes the myopia this causes lead to serious errors of judgement
Hold up your hands people who haven't had serious errors of judgement for your own code? Thought so.
A hacker, on the other hand, does not perform well producing piecework software components based upon pre-set specifications.
So the authors of glibc are not hackers? And the people working on NFS3 support Linux? And what about those who write drivers for new file systems, like NTFS? In all those cases there are pre-set specifications.
hackers generally require access to the artifact's interior and want to understand all parts of the systems they work on, including the components contributed by others.
This is just as true for "workers in programming teams." Or, is a prerequiste of hackers that they cannot work in teams?
the difference between the hacker's approach and those of the industrial programmer is one of outlook: between an agoric, integrated and holistic attitude towards the creation of artifacts and a proprietary, fragmented and reductionist one.
I work in industry. I want to see code. He has asserted a false dichotomy.
Hacking as a method for system development originated as a grass-roots reaction to attempts to impose an industrial mode of production on the development of software.
Here the author confuses the word "Taylorism" (which is now rarely practiced in its original form) with "industrial." Couple of questions: Are the Mozilla developers, who apply pretty standard industry practices (proposals, designs, bug tracking, build management, etc.) hackers? Are software companies which have non-existent internal management full of hackers?
Thanks for the verification :)
if this is the type of stuff on segfault I will certainly be checking it out from now on.
For the most part, no, it isn't.
Here's an expansion of the comments I made on
intellectualcapital.com:
Eric Raymond says there are two styles of software distribution, "closed source" and "open source." He equivalences the latter with "the Internet engineering tradition" and says they are "fundamentally opposed styles of software development." I believe he is wrong. The world is not polarized into two camps; there are many intermediate forms.
Consider a commerical product where customers get the software and are allowed to modify it but not redistribute the source or modifications. This can be loosened someone by allowing patches to be redistributed gratis to other licensed customers. In this case, the customer gets nearly all of the benefits of "open source" (peer review, a fallback path if the supplier stops working on the project, etc.) without the need for granting redistribution rights.
I worked on one proprietary software package which was released with source code under a license similar to that I gave, so this does happen in real life.
I see very, very few additional advantages to the *customer* between this model and the "open source" model advocated by Raymond. I have never heard him address the need for nor benefit of the redistribution requirement.
On the argument he gave, that of a company which just spent several million on closed, propritary software. He forgot to mention that the normal solution to that problem is for the supplier to put the source code in escrow. If the supplier goes out of business or can no longer provide support or revisions, the customer gets the source code. This is standard.
Finally, he omits to mention that some "closed source" software has been found to be of very high quality. There's a very famous study of the quality of the space shuttle avionics code. As I recall, there was one problem found in the integration level, and it wasn't serious enough to affect the mission. The problem is this sort of code review is very, very expensive. I believe it was in the many hundreds of dollars per line of code.
I'll just take it as coincidence that this Slashdot article appeared a few days after my Segfault "news" article on the same subject.