I must apologise for being obtuse. I did complete my Ph.D in computational physics. I did it part time and and it added time to my program. I started a consulting company to fund my living during the process, and ditched that and the "assistantships" in favor of a real salary and benefits.
I worked 9am-5pm. I came home, made dinner for my wife and I, and then wrote my thesis from 7pm till usually 1-2am. Repeated this daily for about two years. Submitted the thesis, defended, passed, voila.
Inside of that time, I have seen my remuneration increase rapidly. But I was not hired for my prowess at modeling, though I have been blessed enough to specify more favorable parameters for my job (due in large part to some stuff I did that made a large customer happy, see my website for details).
My point had been that all such reports about impending shortages need to be taken with a large helping of skepticism. Assume the other person has an agenda and you will be safer. Yeah, it sounds like paranoia, but I cannot tell you how many times I have run into people thinking that they can have a career as a scientist (as I originally had planned on) when there really is no shortfall, but rather a glut of highly educated people.
I wish you luck with your choices. As long as you understand and accept that there are no jobs in physics waiting for you, you are in good shape. I make the very points to potential grad students that you eluded to, but I also caution them to be realistic, and that physics as a job upon graduation is only slightly less probable than winning the state lottery. As long as they have reasonable advisors, a realistic department policy (my department made taking outside classes almost impossible) towards external classes, and lots of industry connections, there shouldn't be too much of a problem in absorbing graduates. Just don't expect to do science.
In the mid 80s, back in college, a document called the "House" report was put out by the NSF detailing the "imminent shortage of scientists and engineers." Of course, since the NSF (the National Science Foundation) put this report out, it had to be correct... right?
Well... no. It was basically a fabrication at best, and perpetuated what is generally called "The Myth" (capitalized as such) by members of the YSN (Young Scientists Network). The Myth was used to justify increased graduate student spending, e.g. more graduate students into Ph.D. programs. It was used to generate more research dollars, so that more work could be done.
In physics, we were pumping out 1400+ Ph.D's per year. Sounds like too few... right?
Well, it turns out that there were only about 150 tenure track jobs opening up each year, and about the same number of industrial jobs. This is what the House report failed to mention, that the reason the supply was dwindling was that demand for the Ph.D scientists was actually quite low. It was simple economics.
The problem at the time was that few of the undergraduate students at the time really knew where to find this information. Few knew that the report put out by the NSF was not worth the paper it was printed on. Few could verify the research in the report, as most didn't have ready access to the sources.
The end result was a glut of scientists and engineers in the market. Too many. Not enough jobs. There is a general belief these days that there were many apochraphal stories floating about how Ph.D.s were driving cabs and what not else. The entrenched establishment of research professors strongly disbelieved that there were problems getting jobs. They pointed to the back of Physics Today and shouted "look at all of those...". This reminded me of when Ronald Reagan called ketchup a vegatable. Most of those jobs in Physics Today were temporary employment. Very few were for tenure tracks. Few were for permanent positions.
You go to college for 14 years and it would be nice if there was some possibility that you could make more than $20k/year starting.
Today the situation is governed by simple economics. There are not that many people going for Ph.D's, not that many people going after post-docs, etc. Now, you need incentives to keep the students in the program, as there are real attractive alternatives to years of mind-expanding indenture.
Face it. In graduate school, as a hard science type, you are an indentured servant. Have no illusions about this. Your purpose in life is to further a professor's career and publication list. Your purpose is not to get a degree, that is an accident if it happens, and largely the professors want you to take your sweet old time about this. You see, you are cheap labor. You are not in a union (this is changing), you are not a professional, and they can pay you under $10k per year to do their work (60-80 hours/week).
You see, I believed the House report. I believed that there would be a shortage of scientists. I believed that the salaries would be high.
Welcome to reality.
I chose to finish my Ph.D part time. That was gruelling and added 2 years onto my time. However, I was paid reasonable wages by my employer. I worked 1/2 as hard at my employer, and got recognition, rewards, raises.
I learned in time. Many of my friends did not. The system chews you up and spits you out.
There is a lesson here, a nice juicy object lesson for anyone wanting to believe these reports of shortages. Assume that they are written by those with a vested interest in keeping a large supply of cheap talent available. Assume they are written by people who are unaware or wish you to be unaware of the real circumstances. And make sure you look at the department of labors job outlook guides.
Epilogue: Several physics departments that I am aware of have lost their supply of new graduate fodder. Moreover, as they have been declining enrollmentwise in the Ph.D. programs, the number of postdocs have decreased as well. Now there are vacancies. In short order, the other part of the law of supply and demand will kick in... they will be forced to raise wages to attract new blood.
And a final note: I note with more than a little bit of black humor that I have been asked to submit a writeup and bio as I appear to be one of the successful graduates of my Alma Mater. They are learning (and in large part due to a change in leadership, to one with a good clue) that they need to market themselves in order to attract new blood.
All things in moderation, with style, and sense
on
Linux Advocacy Hurts
·
· Score: 2
The greatest danger that Linux advocacy has these days is falling into the trap of the Mac advocates. Advocating Linux does not mean slamming the competition, being rude to those who disagree, being disrespectful of others opinions. Advocating Linux means marketing Linux, by telling the world of the positive points, and carefully deflecting the negative points.
The trap that the Mac advocates fell into is that they fell in love with their own conclusions. They (unfortunately) concluded that: a) the mac was a superior hardware platform b) the OS was superior software c) because of a and b, anyone using a mac was superior d) and because of c, anyone not using a mac was using a silly platform.
Well, that was all hogwash, but in their eyes, the non mac users are the unwashed masses. They treat people as such (some very good friends are Mac bigots, and I like them despite their handicap), and basically one has to let them have their say, let their dander settle, and move on.
These days, the vast majority of those who send in such vituperative or highly vitrolic responses to those with negative things to say about Linux, tend to do so without style, without moderation, and often without a clue. They proceed with ad hominem attacks, with levels of bile and disgust that one would not expect from a rational person.
What they miss is that constructive engagement is the correct approach. By kindly and humbly pointing out mistakes, incorrect facts, missed assumptions, etc., you are more likely to have a mind altering affect than with the often thermonuclear email blasts.
I have done the constructive engagement many times, and have seen significant progress in attitudes. Unfortunately the mob mentality could easily destroy this, by exposing those who are critical to sophmoric blasts of flamage.
You dont market a product by yelling at someone and calling them names for being critical.
That is after all what advocacy is all about... marketing. Linux advocacy is marketing the linux products. Don't let it become as bad as Mac advocacy, which borders on rabid fanaticism. That would in the end have the same effect upon linux as it did upon the Mac. It would sink it.
Microsoft has to do this. Systems like Linux are very compelling on the server front, and this is about the only way that they can really compete with Linux. After all, they cannot give their bread and butter away for free now... can they?
Moreover, Intel probably is not terribly serious about this market, as they will sell the server running Linux or running Windows. So this is at best a distraction, and a low profit (e.g. low attention span) direction. In Microsoft's case it is important, as the economics of Linux as a small workgroup server are really hard to beat, especially in direct comparison to NT server.
I suspect that the company that is more concerned about this is Sun. Linux has already munched its low end business and this thin server idea basically nails the coffin on any realistic strategy that they might try as a comeback in this market. I have to admit that this is all getting rather amusing.
When I type quickly, I cannot seem to spell pharmacophore. Oh well..
Actually, my model (guessed at from the concepts I mentioned) is basically a polynomial times a decaying exponential. No exponential growth here, just ordinary impulse and decay. The critical exponents are features of the statistics, and allow you to avoid talking about nested distributions.
I was unaware of that paper. I agree that the binning is unfortunate.
Actually it would be interesting to model the/. effect, and present a paper somewhere:-)
The floppy macromolecule talk was Leslie Kuhn, Mike Thorpe, and a post-doc whos name eludes me at this moment. It was interesting, though not specifically set up for pharmacore identification. It was more in line to examine the action of HIV protiease, and the actual physics of its actions. I thought the analysis was quite striking in terms of being able to determine which parts of the molecule act as rigid bodies, and which parts are flexible. Kind of a mechanical version of SAR. Using this information, they are looking for new methods of attacking/disabling the protiease by enhancing or decreasing flexibility in various regions. I do not know if this has more general pharmacophore applicability, but it is a neat idea.
As for ab initio and other fun things, these guys are trying to convince me that I need to jettison DFT in favor of CCS{D,DT,DTQ,...} in order to get better energetics for my MD codes... Unfortunately an additional seven orders of magnitude in algorithm time doesnt appeal to me. I like to study things with many thousands of basis functions, and something of order 10**30 ((10**3)**10) is kinda hard to calculate... quickly.
If you want to continue this discussion outside this forum, email me at landman@sgi.com.
Well, I just got back from Atlanta and the meeting about an hour ago. It was great. Stephen Hawking spoke on "The Universe in a Nutshell" and I saw many friends, mentors, and even my thesis advisor running around.
As a friend would say, it was a "geek fest". I spent the plane ride back with an acquaintance talking about his research. Funny how much interest there is in this stuff... only if there were... ya know... JOBS in it?
Ok... ok. Most of the stuff talked about would put the most hardy computer geek to sleep. 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). But on the other hand, there were some very cool talks, and a large number of nobel laureates, not to mention some really cool exhibits.
How does any of this tie into the slashdot effect?
In a nutshell, think of the slashdot effect (SE) as an avalanche, or a statistical mechanical critical exponent problem. Once a link is posted, it is only a matter of a short time (tau) that this information gets out, and in short order you get a polynomial in tau times something like exp(-t/tau) people following that link. Thus, the target of the link, with a server that is not configured to deal with recognizing the first derivative of this number of people as a function of time, falling over and coughing uncontrollably until it starts doing things like
dd if=/dev/null of=/dev/hda bs=65536
Properly set up servers frequently check slashdot for existance of links, so they have measurements of tau, and estimates of the polynomial. With this in hand, they can detect a slashdot avalanche, and quitely disconnect the port...
Ok, its been a long day, with one canceled flight out of Atlanta...
is to be sophmoric about someones concerns and questions. To label them a crackpot a-priori without trying to help them.
Patience is the best way to win people over. If you cannot be patient, disengage and shoot your flamage out to/dev/null.
There is nothing wrong with using one's full name, and aside from that if he happened to be proud of the fact that he busted his butt for years after college in order to add a few letters, then the least that should be done is to accord the fellow some respect. Some of us have worked very hard for those few extra letters and we have every right to be proud of them.
He may have made factual errors and drawn incorrect conclusions (I even sent off a nice letter explaining what they were). That is no reason whatsoever to go off the deep end.
The funny thing is that this message is a reply to a message posted by "nyet" which read exactly the way this lawyer fellow complains he got treated here. This is sad folks.
Are/.ers arrogant? Some are. Are the prone to flamage? Put another way, why is Rob implementing 400 moderators? Yes, there is an unfortunate propensity to go thermonuclear. Can this be helped? I hope so, otherwise a big chunk of potential users are going to be alienated by those who aren't able to figure out that flames burn the image of Linux as much as they do the person they are aimed at.
I have to admit that I am biased (in large part due to my employer). I would not consider a dual PII generally a good multiprocessor for numerically intensive codes. This has nothing to do with Intel or Linux, or g77/pgf77/absoft, etc.
What I have found on my codes is that small (actually tiny) problems run well on pentia. But reasonable research sized problems cause it to huff and puff. Machines like the alpha or the R10k (and R12k) kick serious butt on the larger problem sizes. What is just insanely cool is to watch your code (efficiently) use all 32 processors, and get something like a 28-30x speedup.
But, as I said, I am biased.
Back to fortran. Jeff Templon has an excellent page on Linux and Fortran. Better is the big fortran link page. This is really a nice resource and is a nice intro to the general Fortran Market setup by Walt Brainerd. I strongly advise visiting this site if you need to think Fortran.
Ok, now some thoughts. Craig Burley and crew have done a positively bang up job on g77. It is IMO a useful productive research tool... with a caveat or two.
First, it really is just a front end to the gcc back end, so there are many... gcc-isms... floating about.
Second, while optimization is OK, it is generally tied to the gcc optimization, which has traditionally not been very good. The egcs project has had a much better track record on getting real optimization into the compiler. Folks, if your runs can take years, 5% DOES matter. Optimization on pentia is not just -O, you need things like
Third, and most important for me, it (nor egcs) knows nothing whatsoever about multiprocessing.
In short, g77 and egcs in general are awesome tools. But unless you work on small problems, they are not suitable. You will need some better tools, and that involves passing over some money in this case.
I like the Portland Group tools, though the KAI tools are effectively identical to what you use on big supers like Origins. Unfortunately, I do not think KAI supports Linux any longer. Maybe we can all write them a nice letter on how they could drop support for some underused platform for computation (some come to mind here:-) ) in favor of Linux. Market size and all that.
As the author of the referant article wrote, most fortran users want all the speed they can get, so you need to look at what your code spends the most time doing, and figure out if it is doing it the right and most efficient way, or if your system is correctly designed for speed, or if you are hitting one area of your system really hard, and thus causing a bottleneck. In short, if you need to design for speed, start out with a workstation design, and not a PC design. You likely will need massive memory and IO bandwidth to complement an insanely fast CPU. Putting an Alpha into a PC architecture should be considered a capital crime. It makes much more sense to put it into something like a DS20, a T3E or some other design (I can fantasize about an Alpha in an Octane or an Origin, that would be a complete screamer... a memory and IO bus capable of feeding the processor at its full speed... shudder).
The language and its implementation are important, but so is the fundamental system design. You need to avoid bottlenecks everywhere.
Ok, I am biased. I want to get that out of the way.
I have used Linux (in fact I am doing so right now), I have used Solaris.
What matters most of all in the application (database) is an excellent system throughput design. You will not get there by slapping a few disks onto an IDE controller and tossing Linux on. You will get there by careful design of the system, looking for and avoiding potential bottlenecks.
Just because a system is a "PC" doesnt mean it is slow by default, and just because a box is a workstation doesnt mean it is fast. I have seen fast PC boxes positively whup the Solaris boxes I have played with, at a large variety of tasks. But then again, as I said, I am biased, so this news was quite entertaining to me.
If you are building a database, you need a striped file system, or enough controllers and spindles to be able to push data through at an acceptable rate. IDE disks are wonderful... for desktop applications. You need the combination of command tag queuing and disconnection that you get with SCSI or FibreChannel in order to realize performance.
Oracle likes reading and writing to raw partitions, so you might want to look at getting a machine and a set of USCSI controllers (4 controllers ought to be reasonable for a small DB). If you use 10k RPM disks, then you need about 2-3 per controller. The 7.2k RPM disks can go about 4 per controller. If you are the least bit interested in not having this thing be slow, then you will not load up the SCSI channels with disks. This is a very bad thing to do, positively kills performance.
If you are running off of a file system, you will want to have a stripe running across you controllers. 4 channels with 4 disks is a 16 way stripe in the making, and you should be able to push 160 MB/s peak through that with the 7.2kRPM disks. Depending upon the size of the reads and writes, you might be able to get a sizeable fraction of that, though databases tend to do things in tiny 1k or so blocks.
Then you need enough memory for the database to be partially cacheable... I tend to indicate erring on the too large side if at all possible. Any extra ram will initially be used up by disk cache, and subsequently, by an application or 3.
Further, you need to get data in and out of the system, so you probably need multiple ethernets. Now I know Linux can effectively load balance across its physical media, as can IRIX (I told you I was biased), but I dont think Solaris can.
Finally, you have to ask yourself will this database get really large, or will it start requiring massive reorganization of resources to support it... building too small a solution can kill most any project in its infancy, building too large a project... You need to make sure that your solution can scale with the problem requirements. As I said, I am biased. If the solution never has to scale, all of this can be done on Linux. If the solution has to scale, you need scalable hardware. I would look into other vendors than just Sun at that point:-)
As an employee of one of the companies that attended I guess I have an interesting perspective on this, especially with my rather strong advocacy of Linux and Open source in general within my organization. I am not a spokesdroid for my employer, I am here of my own accord and desire, on my own time, and on my own linux machine. And please pardon my spelling and grammar, I dont have a web-ified ispell:-)
I hope that people read past the title, and into this message, we will see if there is a response...
Ok, here is the main idea of this: The OSS model is specifically designed to maximize quality ("all bugs are shallow with many eyes" ESR), and it does this for the collective good of the community. That is, the function of this model is to generally make peoples lives easier, as bugs will be more transparent, easier to spot and fix. Or taking a step back, and slightly restating it... what is important is the community of users and coders, and their ability to get their work done with as little hinderence as possible. This is a great idea, and lo and behold it seems to be working. What matters in the end is not the success or failure of a single user/coder, but generally the stability of the community, and how a particular change may affect not just one or two users, but everyone. The needs of the community are placed ahead of the needs of one coder/user. This is a sort of social communitism (not communism, but a community focus where one has to look beyond themselves).
Unfortunately, this tends to clash with the mindset of the commercial folks (suits as we are called). In the commercial world, what matters in the end is profit. The tools to make a profit involve intellectual property, differentiation, and demand generation. By its nature, it is not designed to be cooperative, but competitive. While some lambast anyone in a "suit", it is important not to be too antagonistic of the other "side", after all, we are trying to grow the community.
So, here we have two different worlds, one with a sense of significant community and mutual betterment, and another community of competitors, used to differentiation and need-to-know type of scenarios. Can they get along? Should they?
In my opinion (not my companies, they dont tell me, and I have to read about their opinion in Yahoo press releases) there is significant benefit to growing the OSS model within an organization. For one, it can strengthen ties tremendously with the community (which will help the company sell future products). Secondly, with many eyes... better products emerge. Just ask netscape. There are other reasons as well, and I would like to see a continuation of this discussion.
So my point (finally he makes one!) is that maybe the community can help the companies develop the internal OSS model that needs to be in place for this to evolve. I am hoping that the community at large feels that the companies can bring value to this process, and will help the process develop. Remember, that the companies need to make money in the end, so part of this process is to help put out a reasonable business model (eek... business school words!) that works with OSS. Once again, we can turn to Netscape, but they are not necessarily the only model. Cygnus, and many others are quite successful. Once again, we need to appeal to the many eyes (or in this case, many brains) to think this one out.
Overall, the report was interesting, and I hope we see more such meetings.
to GNU or not to GNU that is the question
on
Wired on RMS
·
· Score: 1
Folks:
Maybe RMS should have called GNU the RMS system. GNU is so thoroughly embedded within BSD, Linux, and the commercial Unicies, and free software in general (the egcs project, the sendmail code, ssh, etc) that maybe he needs to rethink what he wants.
From reading the article, RMS wants recognition. Well, unless he is looking for a coronation or something like that, shit, he has it! His tools are used widely by tens of millions of people every single day, whether or not he knows it.
IMO what he should do if he wants his ego satiated, is to stop the "GNU/Linux" stuff. It is Linux, and leave that alone. Start getting some people to work with him on spreading the message that GNU makes Linux possible, hell it makes most Unix possible. Talk about this, and talk about his contribution to starting and sheparding GNU. Point to Linux as a successful example of the use and power of the GNU approach, as well as the OSS model, and Free software in general. Stop doing the penis length comparison of "your kernel.c is only 3% of the OS," as a) nobody cares, b) sounds like massively sour grapes there dude.
Talk positively about how he laid the foundation which enabled Linux and xBSD to grow. That should inject him with some serious ego boosters, as people realize that Linus was able to bring about his work in the context of RMS's work. That is a pretty powerful story.
Jesse seems to have missed the point rather badly, which is suprising in that he made the point quite eloquently previously.
That is, you use the right tool for the right job. Jesse seems to miss this point when he recommends MSFT over Linux, as he argues that it really cannot win.
His arguement is chock full of reasoning by incorrect analogies and many other well known logical fallicies. Well, I guess this is to be anticipated, as there is no linux gravy train for journalists...
Seriously, if Jesse wants support for Linux, my company would happily sell it to him, as would a number of others. Single point of contact, multibillion dollar corporation that will not be going away any time soon.
One shouldn't make decisions with their heart, as he did (or wallet?). The decision needs to be based upon sound reasonable practices and information. With that in mind, there is little reason to deploy an NT server, where a less expensive to implement and maintain, and more stable Linux platform would do a better job. The same is true on the desktop.
Nah there's plently of dinky little 3d VRML objects around. You get a pencil or something and you can rotate it with your mouse. Great fun for about 2 minutes.
Funny, most of the VRML I have seen has been quite specific applications. Most of the Java I have seen has been stupid animated icons.
There is a nice jumping place to other VRML worlds, and it is reachable from here. The vast majority of worlds out there are not little play toys, which is not what you can say about the vast majority of java applets.
Somehow, I think you are thinking of java and its ilk when you talk about dinky little 3d animations on a web page.
I wrote a weblication to convert molecular structure files to VRML-1 in order to be able to display them in a platform independent manner. Unfortunately the sorry state of VRML on Linux isnt helping the platform independence, but that is for another thread.
You can see some of the molecules I converted off of my page (click on the molecules button), or more directly off of the quickie weblication page for those with frames/graphics impaired browsers. I used either my tool (genVRML) or the wonderful VMD program which runs great on the SGIs and reasonably on the Linux machines.
I have also seen virtual museum walkthroughs on the web which allow visitors to get a feel for how they want to plan their visits. I have seen virtual astronomy and anatomy labs.
VRML is actually quite a good technology. Unfortunately you need a rather powerful machine with excellent graphics to do a good job at visualizing anything that is moderately complex. Your machine needs a fast OpenGL implementation and hardware texturing. This means once again, an SGI box or a nice Linux box with the accelerated X servers and the hardware stuff with Mesa.
Slashdot Mirror
I must apologise for being obtuse. I did complete my Ph.D in computational physics. I did it part time and and it added time to my program. I started a consulting company to fund my living during the process, and ditched that and the "assistantships" in favor of a real salary and benefits.
I worked 9am-5pm. I came home, made dinner for my wife and I, and then wrote my thesis from 7pm till usually 1-2am. Repeated this daily for about two years. Submitted the thesis, defended, passed, voila.
Inside of that time, I have seen my remuneration increase rapidly. But I was not hired for my prowess at modeling, though I have been blessed enough to specify more favorable parameters for my job (due in large part to some stuff I did that made a large customer happy, see my website for details).
My point had been that all such reports about impending shortages need to be taken with a large helping of skepticism. Assume the other person has an agenda and you will be safer. Yeah, it sounds like paranoia, but I cannot tell you how many times I have run into people thinking that they can have a career as a scientist (as I originally had planned on) when there really is no shortfall, but rather a glut of highly educated people.
I wish you luck with your choices. As long as you understand and accept that there are no jobs in physics waiting for you, you are in good shape. I make the very points to potential grad students that you eluded to, but I also caution them to be realistic, and that physics as a job upon graduation is only slightly less probable than winning the state lottery. As long as they have reasonable advisors, a realistic department policy (my department made taking outside classes almost impossible) towards external classes, and lots of industry connections, there shouldn't be too much of a problem in absorbing graduates. Just don't expect to do science.
In the mid 80s, back in college, a document called the "House" report was put out by the NSF detailing the "imminent shortage of scientists and engineers." Of course, since the NSF (the National Science Foundation) put this report out, it had to be correct... right?
Well... no. It was basically a fabrication at best, and perpetuated what is generally called "The Myth" (capitalized as such) by members of the YSN (Young Scientists Network). The Myth was used to justify increased graduate student spending, e.g. more graduate students into Ph.D. programs. It was used to generate more research dollars, so that more work could be done.
In physics, we were pumping out 1400+ Ph.D's per year. Sounds like too few... right?
Well, it turns out that there were only about 150 tenure track jobs opening up each year, and about the same number of industrial jobs. This is what the House report failed to mention, that the reason the supply was dwindling was that demand for the Ph.D scientists was actually quite low. It was simple economics.
The problem at the time was that few of the undergraduate students at the time really knew where to find this information. Few knew that the report put out by the NSF was not worth the paper it was printed on. Few could verify the research in the report, as most didn't have ready access to the sources.
The end result was a glut of scientists and engineers in the market. Too many. Not enough jobs. There is a general belief these days that there were many apochraphal stories floating about how Ph.D.s were driving cabs and what not else. The entrenched establishment of research professors strongly disbelieved that there were problems getting jobs. They pointed to the back of Physics Today and shouted "look at all of those...". This reminded me of when Ronald Reagan called ketchup a vegatable. Most of those jobs in Physics Today were temporary employment. Very few were for tenure tracks. Few were for permanent positions.
You go to college for 14 years and it would be nice if there was some possibility that you could make more than $20k/year starting.
Today the situation is governed by simple economics. There are not that many people going for Ph.D's, not that many people going after post-docs, etc. Now, you need incentives to keep the students in the program, as there are real attractive alternatives to years of mind-expanding indenture.
Face it. In graduate school, as a hard science type, you are an indentured servant. Have no illusions about this. Your purpose in life is to further a professor's career and publication list. Your purpose is not to get a degree, that is an accident if it happens, and largely the professors want you to take your sweet old time about this. You see, you are cheap labor. You are not in a union (this is changing), you are not a professional, and they can pay you under $10k per year to do their work (60-80 hours/week).
You see, I believed the House report. I believed that there would be a shortage of scientists. I believed that the salaries would be high.
Welcome to reality.
I chose to finish my Ph.D part time. That was gruelling and added 2 years onto my time. However, I was paid reasonable wages by my employer. I worked 1/2 as hard at my employer, and got recognition, rewards, raises.
I learned in time. Many of my friends did not. The system chews you up and spits you out.
There is a lesson here, a nice juicy object lesson for anyone wanting to believe these reports of shortages. Assume that they are written by those with a vested interest in keeping a large supply of cheap talent available. Assume they are written by people who are unaware or wish you to be unaware of the real circumstances. And make sure you look at the department of labors job outlook guides.
Epilogue: Several physics departments that I am aware of have lost their supply of new graduate fodder. Moreover, as they have been declining enrollmentwise in the Ph.D. programs, the number of postdocs have decreased as well. Now there are vacancies. In short order, the other part of the law of supply and demand will kick in... they will be forced to raise wages to attract new blood.
And a final note: I note with more than a little bit of black humor that I have been asked to submit a writeup and bio as I appear to be one of the successful graduates of my Alma Mater. They are learning (and in large part due to a change in leadership, to one with a good clue) that they need to market themselves in order to attract new blood.
The greatest danger that Linux advocacy has these days is falling into the trap of the Mac advocates. Advocating Linux does not mean slamming the competition, being rude to those who disagree, being disrespectful of others opinions. Advocating Linux means marketing Linux, by telling the world of the positive points, and carefully deflecting the negative points.
The trap that the Mac advocates fell into is that they fell in love with their own conclusions. They (unfortunately) concluded that:
a) the mac was a superior hardware platform
b) the OS was superior software
c) because of a and b, anyone using a mac was superior
d) and because of c, anyone not using a mac was using a silly platform.
Well, that was all hogwash, but in their eyes, the non mac users are the unwashed masses. They treat people as such (some very good friends are Mac bigots, and I like them despite their handicap), and basically one has to let them have their say, let their dander settle, and move on.
These days, the vast majority of those who send in such vituperative or highly vitrolic responses to those with negative things to say about Linux, tend to do so without style, without moderation, and often without a clue. They proceed with ad hominem attacks, with levels of bile and disgust that one would not expect from a rational person.
What they miss is that constructive engagement is the correct approach. By kindly and humbly pointing out mistakes, incorrect facts, missed assumptions, etc., you are more likely to have a mind altering affect than with the often thermonuclear email blasts.
I have done the constructive engagement many times, and have seen significant progress in attitudes. Unfortunately the mob mentality could easily destroy this, by exposing those who are critical to sophmoric blasts of flamage.
You dont market a product by yelling at someone and calling them names for being critical.
That is after all what advocacy is all about... marketing. Linux advocacy is marketing the linux products. Don't let it become as bad as Mac advocacy, which borders on rabid fanaticism. That would in the end have the same effect upon linux as it did upon the Mac. It would sink it.
Microsoft has to do this. Systems like Linux are very compelling on the server front, and this is about the only way that they can really compete with Linux. After all, they cannot give their bread and butter away for free now... can they?
Moreover, Intel probably is not terribly serious about this market, as they will sell the server running Linux or running Windows. So this is at best a distraction, and a low profit (e.g. low attention span) direction. In Microsoft's case it is important, as the economics of Linux as a small workgroup server are really hard to beat, especially in direct comparison to NT server.
I suspect that the company that is more concerned about this is Sun. Linux has already munched its low end business and this thin server idea basically nails the coffin on any realistic strategy that they might try as a comeback in this market. I have to admit that this is all getting rather amusing.
When I type quickly, I cannot seem to spell pharmacophore. Oh well..
Actually, my model (guessed at from the concepts I mentioned) is basically a polynomial times a decaying exponential. No exponential growth here, just ordinary impulse and decay. The critical exponents are features of the statistics, and allow you to avoid talking about nested distributions.
I was unaware of that paper. I agree that the binning is unfortunate.
Joe
Aha... I think I know you :-)
/. effect, and present a paper somewhere :-)
... quickly.
Actually it would be interesting to model the
The floppy macromolecule talk was Leslie Kuhn, Mike Thorpe, and a post-doc whos name eludes me at this moment. It was interesting, though not specifically set up for pharmacore identification. It was more in line to examine the action of HIV protiease, and the actual physics of its actions. I thought the analysis was quite striking in terms of being able to determine which parts of the molecule act as rigid bodies, and which parts are flexible. Kind of a mechanical version of SAR. Using this information, they are looking for new methods of attacking/disabling the protiease by enhancing or decreasing flexibility in various regions. I do not know if this has more general pharmacophore applicability, but it is a neat idea.
As for ab initio and other fun things, these guys are trying to convince me that I need to jettison DFT in favor of CCS{D,DT,DTQ,...} in order to get better energetics for my MD codes... Unfortunately an additional seven orders of magnitude in algorithm time doesnt appeal to me. I like to study things with many thousands of basis functions, and something of order 10**30 ((10**3)**10) is kinda hard to calculate
If you want to continue this discussion outside this forum, email me at landman@sgi.com.
Joe
Well, I just got back from Atlanta and the meeting about an hour ago. It was great. Stephen Hawking spoke on "The Universe in a Nutshell" and I saw many friends, mentors, and even my thesis advisor running around.
As a friend would say, it was a "geek fest". I spent the plane ride back with an acquaintance talking about his research. Funny how much interest there is in this stuff... only if there were... ya know... JOBS in it?
Ok... ok. Most of the stuff talked about would put the most hardy computer geek to sleep. 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). But on the other hand, there were some very cool talks, and a large number of nobel laureates, not to mention some really cool exhibits.
How does any of this tie into the slashdot effect?
In a nutshell, think of the slashdot effect (SE) as an avalanche, or a statistical mechanical critical exponent problem. Once a link is posted, it is only a matter of a short time (tau) that this information gets out, and in short order you get a polynomial in tau times something like exp(-t/tau) people following that link. Thus, the target of the link, with a server that is not configured to deal with recognizing the first derivative of this number of people as a function of time, falling over and coughing uncontrollably until it starts doing things like
dd if=/dev/null of=/dev/hda bs=65536
Properly set up servers frequently check slashdot for existance of links, so they have measurements of tau, and estimates of the polynomial. With this in hand, they can detect a slashdot avalanche, and quitely disconnect the port...
Ok, its been a long day, with one canceled flight out of Atlanta...
is to be sophmoric about someones concerns and questions. To label them a crackpot a-priori without trying to help them.
/dev/null.
/.ers arrogant? Some are. Are the prone to flamage? Put another way, why is Rob implementing 400 moderators? Yes, there is an unfortunate propensity to go thermonuclear. Can this be helped? I hope so, otherwise a big chunk of potential users are going to be alienated by those who aren't able to figure out that flames burn the image of Linux as much as they do the person they are aimed at.
Patience is the best way to win people over. If you cannot be patient, disengage and shoot your flamage out to
There is nothing wrong with using one's full name, and aside from that if he happened to be proud of the fact that he busted his butt for years after college in order to add a few letters, then the least that should be done is to accord the fellow some respect. Some of us have worked very hard for those few extra letters and we have every right to be proud of them.
He may have made factual errors and drawn incorrect conclusions (I even sent off a nice letter explaining what they were). That is no reason whatsoever to go off the deep end.
The funny thing is that this message is a reply to a message posted by "nyet" which read exactly the way this lawyer fellow complains he got treated here. This is sad folks.
Are
Joe
I have to admit that I am biased (in large part due to my employer). I would not consider a dual PII generally a good multiprocessor for numerically intensive codes. This has nothing to do with Intel or Linux, or g77/pgf77/absoft, etc.
:-) ) in favor of Linux. Market size and all that.
What I have found on my codes is that small (actually tiny) problems run well on pentia. But reasonable research sized problems cause it to huff and puff. Machines like the alpha or the R10k (and R12k) kick serious butt on the larger problem sizes. What is just insanely cool is to watch your code (efficiently) use all 32 processors, and get something like a 28-30x speedup.
But, as I said, I am biased.
Back to fortran. Jeff Templon has an excellent page on Linux and Fortran. Better is the big fortran link page. This is really a nice resource and is a nice intro to the general Fortran Market setup by Walt Brainerd. I strongly advise visiting this site if you need to think Fortran.
Ok, now some thoughts. Craig Burley and crew have done a positively bang up job on g77. It is IMO a useful productive research tool... with a caveat or two.
First, it really is just a front end to the gcc back end, so there are many... gcc-isms... floating about.
Second, while optimization is OK, it is generally tied to the gcc optimization, which has traditionally not been very good. The egcs project has had a much better track record on getting real optimization into the compiler. Folks, if your runs can take years, 5% DOES matter. Optimization on pentia is not just -O, you need things like
-O3 -malign-double -malign-functions=2 -funroll-loops -ffast-math
among others for decent performance.
Third, and most important for me, it (nor egcs) knows nothing whatsoever about multiprocessing.
In short, g77 and egcs in general are awesome tools. But unless you work on small problems, they are not suitable. You will need some better tools, and that involves passing over some money in this case.
I like the Portland Group tools, though the KAI tools are effectively identical to what you use on big supers like Origins. Unfortunately, I do not think KAI supports Linux any longer. Maybe we can all write them a nice letter on how they could drop support for some underused platform for computation (some come to mind here
As the author of the referant article wrote, most fortran users want all the speed they can get, so you need to look at what your code spends the most time doing, and figure out if it is doing it the right and most efficient way, or if your system is correctly designed for speed, or if you are hitting one area of your system really hard, and thus causing a bottleneck. In short, if you need to design for speed, start out with a workstation design, and not a PC design. You likely will need massive memory and IO bandwidth to complement an insanely fast CPU. Putting an Alpha into a PC architecture should be considered a capital crime. It makes much more sense to put it into something like a DS20, a T3E or some other design (I can fantasize about an Alpha in an Octane or an Origin, that would be a complete screamer... a memory and IO bus capable of feeding the processor at its full speed... shudder).
The language and its implementation are important, but so is the fundamental system design. You need to avoid bottlenecks everywhere.
Joe
Ok, I am biased. I want to get that out of the way.
... You need to make sure that your solution can scale with the problem requirements. As I said, I am biased. If the solution never has to scale, all of this can be done on Linux. If the solution has to scale, you need scalable hardware. I would look into other vendors than just Sun at that point :-)
I have used Linux (in fact I am doing so right now), I have used Solaris.
What matters most of all in the application (database) is an excellent system throughput design. You will not get there by slapping a few disks onto an IDE controller and tossing Linux on. You will get there by careful design of the system, looking for and avoiding potential bottlenecks.
Just because a system is a "PC" doesnt mean it is slow by default, and just because a box is a workstation doesnt mean it is fast. I have seen fast PC boxes positively whup the Solaris boxes I have played with, at a large variety of tasks. But then again, as I said, I am biased, so this news was quite entertaining to me.
If you are building a database, you need a striped file system, or enough controllers and spindles to be able to push data through at an acceptable rate. IDE disks are wonderful... for desktop applications. You need the combination of command tag queuing and disconnection that you get with SCSI or FibreChannel in order to realize performance.
Oracle likes reading and writing to raw partitions, so you might want to look at getting a machine and a set of USCSI controllers (4 controllers ought to be reasonable for a small DB). If you use 10k RPM disks, then you need about 2-3 per controller. The 7.2k RPM disks can go about 4 per controller. If you are the least bit interested in not having this thing be slow, then you will not load up the SCSI channels with disks. This is a very bad thing to do, positively kills performance.
If you are running off of a file system, you will want to have a stripe running across you controllers. 4 channels with 4 disks is a 16 way stripe in the making, and you should be able to push 160 MB/s peak through that with the 7.2kRPM disks. Depending upon the size of the reads and writes, you might be able to get a sizeable fraction of that, though databases tend to do things in tiny 1k or so blocks.
Then you need enough memory for the database to be partially cacheable... I tend to indicate erring on the too large side if at all possible. Any extra ram will initially be used up by disk cache, and subsequently, by an application or 3.
Further, you need to get data in and out of the system, so you probably need multiple ethernets. Now I know Linux can effectively load balance across its physical media, as can IRIX (I told you I was biased), but I dont think Solaris can.
Finally, you have to ask yourself will this database get really large, or will it start requiring massive reorganization of resources to support it... building too small a solution can kill most any project in its infancy, building too large a project
As an employee of one of the companies that attended I guess I have an interesting perspective on this, especially with my rather strong advocacy of Linux and Open source in general within my organization. I am not a spokesdroid for my employer, I am here of my own accord and desire, on my own time, and on my own linux machine. And please pardon my spelling and grammar, I dont have a web-ified ispell :-)
I hope that people read past the title, and into this message, we will see if there is a response...
Ok, here is the main idea of this: The OSS model is specifically designed to maximize quality ("all bugs are shallow with many eyes" ESR), and it does this for the collective good of the community. That is, the function of this model is to generally make peoples lives easier, as bugs will be more transparent, easier to spot and fix. Or taking a step back, and slightly restating it... what is important is the community of users and coders, and their ability to get their work done with as little hinderence as possible. This is a great idea, and lo and behold it seems to be working. What matters in the end is not the success or failure of a single user/coder, but generally the stability of the community, and how a particular change may affect not just one or two users, but everyone. The needs of the community are placed ahead of the needs of one coder/user. This is a sort of social communitism (not communism, but a community focus where one has to look beyond themselves).
Unfortunately, this tends to clash with the mindset of the commercial folks (suits as we are called). In the commercial world, what matters in the end is profit. The tools to make a profit involve intellectual property, differentiation, and demand generation. By its nature, it is not designed to be cooperative, but competitive. While some lambast anyone in a "suit", it is important not to be too antagonistic of the other "side", after all, we are trying to grow the community.
So, here we have two different worlds, one with a sense of significant community and mutual betterment, and another community of competitors, used to differentiation and need-to-know type of scenarios. Can they get along? Should they?
In my opinion (not my companies, they dont tell me, and I have to read about their opinion in Yahoo press releases) there is significant benefit to growing the OSS model within an organization. For one, it can strengthen ties tremendously with the community (which will help the company sell future products). Secondly, with many eyes... better products emerge. Just ask netscape. There are other reasons as well, and I would like to see a continuation of this discussion.
So my point (finally he makes one!) is that maybe the community can help the companies develop the internal OSS model that needs to be in place for this to evolve. I am hoping that the community at large feels that the companies can bring value to this process, and will help the process develop. Remember, that the companies need to make money in the end, so part of this process is to help put out a reasonable business model (eek... business school words!) that works with OSS. Once again, we can turn to Netscape, but they are not necessarily the only model. Cygnus, and many others are quite successful. Once again, we need to appeal to the many eyes (or in this case, many brains) to think this one out.
Overall, the report was interesting, and I hope we see more such meetings.
Folks:
Maybe RMS should have called GNU the RMS system. GNU is so thoroughly embedded within BSD, Linux, and the commercial Unicies, and free software in general (the egcs project, the sendmail code, ssh, etc) that maybe he needs to rethink what he wants.
From reading the article, RMS wants recognition. Well, unless he is looking for a coronation or something like that, shit, he has it! His tools are used widely by tens of millions of people every single day, whether or not he knows it.
IMO what he should do if he wants his ego satiated, is to stop the "GNU/Linux" stuff. It is Linux, and leave that alone. Start getting some people to work with him on spreading the message that GNU makes Linux possible, hell it makes most Unix possible. Talk about this, and talk about his contribution to starting and sheparding GNU. Point to Linux as a successful example of the use and power of the GNU approach, as well as the OSS model, and Free software in general. Stop doing the penis length comparison of "your kernel.c is only 3% of the OS," as a) nobody cares, b) sounds like massively sour grapes there dude.
Talk positively about how he laid the foundation which enabled Linux and xBSD to grow. That should inject him with some serious ego boosters, as people realize that Linus was able to bring about his work in the context of RMS's work. That is a pretty powerful story.
Jesse seems to have missed the point rather badly, which is suprising in that he made the point quite eloquently previously.
That is, you use the right tool for the right job. Jesse seems to miss this point when he recommends MSFT over Linux, as he argues that it really cannot win.
His arguement is chock full of reasoning by incorrect analogies and many other well known logical fallicies. Well, I guess this is to be anticipated, as there is no linux gravy train for journalists...
Seriously, if Jesse wants support for Linux, my company would happily sell it to him, as would a number of others. Single point of contact, multibillion dollar corporation that will not be going away any time soon.
One shouldn't make decisions with their heart, as he did (or wallet?). The decision needs to be based upon sound reasonable practices and information. With that in mind, there is little reason to deploy an NT server, where a less expensive to implement and maintain, and more stable Linux platform would do a better job. The same is true on the desktop.
Joe
Nah there's plently of dinky little 3d VRML objects around. You get a pencil or something and you can rotate it with your mouse. Great fun for about 2 minutes.
Funny, most of the VRML I have seen has been quite specific applications. Most of the Java I have seen has been stupid animated icons.
There is a nice jumping place to other VRML worlds, and it is reachable from here. The vast majority of worlds out there are not little play toys, which is not what you can say about the vast majority of java applets.
Somehow, I think you are thinking of java and its ilk when you talk about dinky little 3d animations on a web page.
I wrote a weblication to convert molecular structure files to VRML-1 in order to be able to display them in a platform independent manner. Unfortunately the sorry state of VRML on Linux isnt helping the platform independence, but that is for another thread.
You can see some of the molecules I converted off of my page (click on the molecules button), or more directly off of the quickie weblication page for those with frames/graphics impaired browsers. I used either my tool (genVRML) or the wonderful VMD program which runs great on the SGIs and reasonably on the Linux machines.
I have also seen virtual museum walkthroughs on the web which allow visitors to get a feel for how they want to plan their visits. I have seen virtual astronomy and anatomy labs.
VRML is actually quite a good technology. Unfortunately you need a rather powerful machine with excellent graphics to do a good job at visualizing anything that is moderately complex. Your machine needs a fast OpenGL implementation and hardware texturing. This means once again, an SGI box or a nice Linux box with the accelerated X servers and the hardware stuff with Mesa.