* relaytool is C-specific, and it would be impractical to establish something similar for all the languages and ABIs out there. Instead, I would recommend 2nd level dependencies (i.e., package X requires Y to have been built with Z) to deal with ABI and even API shifts. I recognize that this complicates binary distribution; am I thinking too Gentoo-ish?
* Instead of suggesting to the user to install recommended package Y for primary package X, perhaps his autopackage UI can ask him which recommended packages to install, at the time he is installing package X. Then, the autopackage database can keep track of the his prefs for package X, and resolve forward and reverse Y dependencies transparently.
autopackage's dependency policy concerns me. They argue that rapidly changing deps should be incorporated into a package's codebase, and that conditional deps should be handled at runtime with dlopen() or using their C(++) specific "relaytool".
Two questions:
* Even stable packages can have unstable APIs, e.g. GNOME with their timed release cycles. GNOME apps not distributed with GNOME rely on detailed dependency resolution for safe operation. Shouldn't the automation of this be the primary goal of a tool like autopackage (or emerge, urpmi, etc. etc.)?
* How can the user tweak the functionality, i.e. does he have to manually install a conditional dep, so that some functionality magically starts working in his package of interest? For example, I install evolution, but to get Palm-syncing capability, I have to manually install the gnome-conduits package? This requires extra knowledge in the user, or some notification from the package of interest.
I'm working on my PhD, and teach. What has worked for me:
* Demos, demos, demos. The louder, brighter and more mysterious-seeming, the better.
* Some students are into technology, others are into cosmology and exotic topics. Draw connections between their lives and physics, esp. the possibilities stemming from new developments.
* Be very crisp in your own treatment, so the students see the beauty through complication.
You are not going to achieve social engineering through physics. The goal is to give bright students interested in science something to think about, and hopefully excite their imaginations if they are so inclined.
Also, the table of contents looks like a sea of text. Make the chapter headings much, much larger than the section titles. You already did a good job with the "part" demarcations. This lets the reader see the material in a more hierarchal manner.
You might want to try page-by-section rather than page-by-chapter, and keep the sections short enough so the browser window needn't be scrolled. Powerpoint presentations follow the same rule. (Maybe, leave page-by-chapter as an option if the text is used for reference instead of initial exposure).
Leave references until the end of a part or until the end of the book -- they are not as critical in a textbook as they are in a paper.
Maybe forcibly use larger fonts -- not all users will be savvy enough to increase the size of their default fonts.
Your site is a little annoying with the big fat headline graphics and the blinking.
I used to use GNOME, then switched to Fluxbox, then back to GNOME (v.2, with Epiphany). I think the HIG decisions make it clean (like Fluxbox and co.) but featureful like KDE (which, I agree, is way too cartoon-like).
I don't understand all the whining about Epiphany -- it'll save the session if it crashes, and the session saving on GNOME exit will work when it's stable. I use Epiphany for all my web browsing, including "mission-critical" stuff like paying my bills.
Could IBM pose a challenge to Apple in the notebook market with PPC-driven machines? Since the new PPC chip runs cooler while drawing less power, it might fit the bill. Perhaps someone more familiar with PPC architecture could discuss the technical viability of such a beast.
As for economic/consumer viability, right now nearly all the software I use is source-available (currently through Gentoo, on my Compaq Intel notebook). Nevertheless, iin the future, if I need to use pre-compiled, 3rd party software like Mathematica, IDL, etc. PPC+Linux might prove to be too small a market even with IBM's backing -- vendor "lock-out."
One of my advisors in undergrad, Dan Kleppner, used to write a column called "Reference Frame" for _Physics Today_. In one column, he suggested granting tenure in 10-year terms; renewal might require review by the dean, etc. The idea is to allow academic freedom while preventing indolence.
It was before my time, so I don't know it was received, but I imagined it created some waves -- he's rather famous and well-respected, and has been around the block. He argues his position incisively and eloquently, befitting an English major at Williams.:-)
The highways are a natural evolution from railroads, which are more expensive to build and maintain, and are obviously not as flexible. Demand was, and continues to be, intense.
ARPANET was a defense initiative so major universities that subsisted on defense funding could communicate with one another quickly during the Cold War. I think the rise of the public Internet was a pleasant surprise for everybody.
It's not clear what the return on investment will be for a legislated rollout of broadband. In Canada, you are subsidizing broadband with your heavy taxes, as in Korea. If you read the article, broadband providers are hard-up for profits despite the high population density. How are Canadian providers doing?
You can make the chicken-egg argument that e-tailing, etc. would be more effective if more people had broadband. However, in my estimation, the issue was a lack of demand from those who had access, not a lack of access per se.
I think such an argument would be proven right for the _next_ generation, when today's kids become tomorrow's broadband consumers. But, by then, Korea-style broadband might be prevalent w/o gov't intervention, and the attendant inefficiencies.
If the country as a whole is moving up the foodchain, you need to move with it. If the software industry is outsourcing everything to India, be the person who decides what gets outsourced where.
[...] studies are now needed on the role of exercise in weight loss [...]
Gosh, I wonder what they'll find?
Yeah, I know that there are complications due to gender, body type, genetics, etc. but if you are so fat that an unbalanced, low-cal diet is good for you, then a little exercise can't hurt.
As an American who grew up near the nation's capital, I hold the opposite view. The Concorde is cutting edge military technology misappropriated to a non-existent commercial market. On the other hand, the SR-71 is beautiful because it is purpose-built and every aspect of its design shouts out this purpose. The Blackbird's sole aim is reconnaissance, accomplished by flying higher and faster than anything coming from the ground.
Satellites obviate the SR-71, but for ~20 years it was an archangel.
... "mass piracy" is the object of utility, not the object of a transitive "threatens."
This is why Latin is superior for mechanistic parsing, because it would be the "ablative of utility" (or maybe the dative). These would be different cases than the accusative, which is usually the case for objects of transitive verbs in the active voice in the dead language.
Nevertheless, English flows well in speech and written text more or less, and has become the lingua franca for the Internet Age. Interesting stuff.
Yeah, this post is offtopic, but so is the parent. And I didn't find it funny, either.
When I think of a planet, I think of a fluid (hot non-plasma gas, liquid metal) held together by self-gravity. Contrast with an asteroid, which is a rigid object (i.e., a rock). Furthermore, a planet must orbit around the sun, and not another body; if it does, it should be called a moon.
Whatever. There simply isn't any value added by any of these distributions.
Isn't that application dependent?
I am a power user, and I am using Gentoo here on my laptop -- like butter.
At the physics cluster here, they use Red Hat 8.0, because they need a set of standardized binaries, and didn't want to invest the time to work out the kinks themselves.
At my group back in undergrad, the PhDs had zero time for administration, so they installed Debian and never looked back.
(etc....)
What matters is that a distro a) does what it promises and b) is interoperable with other distros.
Requirement (b) is already handled by tarballs for most distros, and also in some by the low overhead in creating packages for them (e.g., Gentoo).
Requirement (a) is really what separates the distros. We'll see how Yoper fares. If two distros occupy the same niche, the better engineered one will win (well, hype not withstanding...).
Whatever. There simply isn't any value added by any of these distributions.
Isn't that application dependent?
I am a power user, and I am using Gentoo here on my laptop -- like butter.
At the physics cluste here, they use Red Hat 8.0, because they need a set of standardized binaries, and didnn't want to invest the time to work out the kinks themselves.
At my group back in undergrad, the PhDs had zero time for administration, so they installed Debian and never looked back.
(etc....)
What matters is that a distro a) does what it promises and b) is interoperable with other distros.
Requirement (b) is already handled by tarballs for most distros, and also in some by the low overhead in creating packages for them (e.g., Gentoo).
Requirement (a) is really what separates the distros.
Your complaints have merit, esp. the cost. Luckily, I had the fortune of entering an IDL work environment that already had a lot of these issues taken care of (a gov't atmospheric physics group), via wrapper code, if necessary. To address some of these issues specifically (viz. IDL 5.4):
Off-by-one pixel -- I have not encountered this when being careful to use cubic splines; perhaps I am mistaken, since my data sets were huge and with small changes element-to-element. Obviously you have to take care of end cases, and there's no uniform way to do that.
FOR loop limits -- Just use a long integer as your index, like in C(++).
Structs syntax -- Just pretend they're C structures, and regard the hash-type functionality as something that should be used sparingly.
module isolation -- This is a fundamental design drawback from the days of yore, but it would break too much code to fix it now. Python, Ruby, Perl, etc. all handle this much better, obviously; but, how many interpreted specialty languages want to make this overhead-v.-robustness tradeoff?
All things considered, I hope PDL, Python + extensions or Octave prove to be viable alternatives. As another insightful poster wrote, inertia/legacy is the biggest factor in deciding what to use. Some people use C/Fortran because it's the lowest common denominator, but the cure is worse than the disease. Good design, internally and in the user interface, is the way to go.
Now, when will there be a viable free alternative to Mathematia? Sorry, Maxima just doesn't cut it.
I've just built it successfully on my Gentoo laptop (everyone seems to rave about, so I thought I'd try it out). Ostensibly, whatever voodoo the ebuild script does works. Here it is:
I've used a number of different packages, with varying results. Here are my thoughts:
Mathematica -- It's good at everything: symbolic computation, statistical analysis (esp. if you deal in intricate error propagation, where symbolic computation is handy), visualization (w/ some tweaking), and even number-crunching. Has a fantastic built-in library. However, it is a blackbox solution (and I have encountered errors in the past), is awfully slow (can be sped up by accessing the kernel directly through C) and closed source.
IDL -- Great for crunching through large amounts of data for the end-user because it has optimized implicit array math. It has an extensive built=library and is good at producing visualizations. Drawbacks are: blackbox (though it uses well known algorithms out of Numerical Recipes, for example), closed-source, and runs best on Windoze, and has an arcane syntax which is some bastard child of Pascal, Fortran and Perl, but not too bad when you get used to it!
Maple -- Has all of Mathematica's weaknesses but cannot match its built-in capabilities (plotting, extensive symbolic library, statistics, numerical analysis).
Matlab -- Only suitable for numerical computation, and is neither as easy to use nor as replete as IDL.
Are there any viable open-source solutions to either Mathematica or IDL?
CAVEAT: I'm of India descent, but I don't live in India, and the last time I was there was 1996.
I'd say most of the people who can read and write, and certainly those wealthy enough to have access to a computer, know English.
Local language support (Hindi is one, don't forget the other major languages like Bengali, Gujarati, Telugu, etc.) seems more appropriate when Linux usage extends beyond gov't/academia to home and commercial situations.
I skeptical that social change (i.e., adoption of computers/internet across the population) can be effected by supply-side pressure when there are such high barriers to adoption....
Hey, you're doing far more than the vast majority of Americans. Four inches off the waistline in 6 months is damn good. And there's nothing evil about beer or red meat per se.
What I put on the post to which you replied are guidelines, not dogma (hence my wistful surprise when I saw it get modded as flamebait!). I certainly indulge in cheese and crackers and sweets every now and then, and I don't limit myself vis-a-vis my nutrional regimen when I go out on engagements on Friday night or over the weekend. I find that my diet is pretty easy to stick to since a) it tastes good with a little planning and b) I feel better eating healthfully. I love a 3-inch think filet mignon, but tops once/month. When I get a hankering for red meat, I stick to lean cuts or lean ground...
Happiness is the ultimate goal, no? I think health officials try to hard, and turn off people. I also think people try to hard, and turn off themselves. Very few people find true satisfaction in torturing themselves in a game of diminishing returns. I think I've found a comfortable critical point, as have you.
How does one get ripped? Look at all the shredded people, and see what they do and what they have going for them in the environment and genetically. My subjective observations, based on the research I've done to formulate my own diet/training program:
Eat a diet low in saturated fats (generally, animal fats) and simple/refined carbs, high in protein, vegetables and fiber, with just the right amount of complex carbs and essential fatty acids (generally, canola and fish).
Hit the weights. The extra lean muscle mass increases your base metabolism.
Cardio is good, but overrated. It compels your body to raid sugar stores instead of burn fat because the rate in energy expenditure is too high to burn fat efficiently. Having a higher base metabolism is the best strategy since it burns all day.
Eat all day, in small amounts -- increases utilization since your body expects food to be coming in short order. The flip side is if you miss a meal (e.g., if you're traveling), you feel like you want to go into "standby."
It helps to be a good athlete who can pack on muscle easily.
Sleep and relax like you don't have a care in the world -- stress (read: cortisol) is the enemy of looking and feeling healthy.
My own results have been mixed. I got pretty lean late last year when I had time to do things right, and my strength and endurance were quite good, but I didn't gain as much muscle mass as I wanted. I was probably overtraining, lifting four days a week an hour at a time, all out.
This dude is hardcore -- he's probably the top male fitness model out there right now. The only modification I've made is that I lift more and play basketball and do less cardio, and try to eat big after a workout to replenish my muscles.
Slashdot Mirror
Criticisms:
* relaytool is C-specific, and it would be impractical to establish something similar for all the languages and ABIs out there. Instead, I would recommend 2nd level dependencies (i.e., package X requires Y to have been built with Z) to deal with ABI and even API shifts. I recognize that this complicates binary distribution; am I thinking too Gentoo-ish?
* Instead of suggesting to the user to install recommended package Y for primary package X, perhaps his autopackage UI can ask him which recommended packages to install, at the time he is installing package X. Then, the autopackage database can keep track of the his prefs for package X, and resolve forward and reverse Y dependencies transparently.
autopackage's dependency policy concerns me. They argue that rapidly changing deps should be incorporated into a package's codebase, and that conditional deps should be handled at runtime with dlopen() or using their C(++) specific "relaytool".
Two questions:
* Even stable packages can have unstable APIs, e.g. GNOME with their timed release cycles. GNOME apps not distributed with GNOME rely on detailed dependency resolution for safe operation. Shouldn't the automation of this be the primary goal of a tool like autopackage (or emerge, urpmi, etc. etc.)?
* How can the user tweak the functionality, i.e. does he have to manually install a conditional dep, so that some functionality magically starts working in his package of interest? For example, I install evolution, but to get Palm-syncing capability, I have to manually install the gnome-conduits package? This requires extra knowledge in the user, or some notification from the package of interest.
I'm working on my PhD, and teach. What has worked for me:
* Demos, demos, demos. The louder, brighter and more mysterious-seeming, the better.
* Some students are into technology, others are into cosmology and exotic topics. Draw connections between their lives and physics, esp. the possibilities stemming from new developments.
* Be very crisp in your own treatment, so the students see the beauty through complication.
You are not going to achieve social engineering through physics. The goal is to give bright students interested in science something to think about, and hopefully excite their imaginations if they are so inclined.
I would add that, if you are using python, you can get cross-GUI portability with wxPython.
Python won't solve the dependencies problem -- that will have to be done at the OS level, using atp, urpmi, Portage, etc.
You need an index.
Also, the table of contents looks like a sea of text. Make the chapter headings much, much larger than the section titles. You already did a good job with the "part" demarcations. This lets the reader see the material in a more hierarchal manner.
You might want to try page-by-section rather than page-by-chapter, and keep the sections short enough so the browser window needn't be scrolled. Powerpoint presentations follow the same rule. (Maybe, leave page-by-chapter as an option if the text is used for reference instead of initial exposure).
Leave references until the end of a part or until the end of the book -- they are not as critical in a textbook as they are in a paper.
Maybe forcibly use larger fonts -- not all users will be savvy enough to increase the size of their default fonts.
Your site is a little annoying with the big fat headline graphics and the blinking.
I used to use GNOME, then switched to Fluxbox, then back to GNOME (v.2, with Epiphany). I think the HIG decisions make it clean (like Fluxbox and co.) but featureful like KDE (which, I agree, is way too cartoon-like).
I don't understand all the whining about Epiphany -- it'll save the session if it crashes, and the session saving on GNOME exit will work when it's stable. I use Epiphany for all my web browsing, including "mission-critical" stuff like paying my bills.
Could IBM pose a challenge to Apple in the notebook market with PPC-driven machines? Since the new PPC chip runs cooler while drawing less power, it might fit the bill. Perhaps someone more familiar with PPC architecture could discuss the technical viability of such a beast.
As for economic/consumer viability, right now nearly all the software I use is source-available (currently through Gentoo, on my Compaq Intel notebook). Nevertheless, iin the future, if I need to use pre-compiled, 3rd party software like Mathematica, IDL, etc. PPC+Linux might prove to be too small a market even with IBM's backing -- vendor "lock-out."
One of my advisors in undergrad, Dan Kleppner, used to write a column called "Reference Frame" for _Physics Today_. In one column, he suggested granting tenure in 10-year terms; renewal might require review by the dean, etc. The idea is to allow academic freedom while preventing indolence.
:-)
It was before my time, so I don't know it was received, but I imagined it created some waves -- he's rather famous and well-respected, and has been around the block. He argues his position incisively and eloquently, befitting an English major at Williams.
I think they made it up
The highways are a natural evolution from railroads, which are more expensive to build and maintain, and are obviously not as flexible. Demand was, and continues to be, intense.
ARPANET was a defense initiative so major universities that subsisted on defense funding could communicate with one another quickly during the Cold War. I think the rise of the public Internet was a pleasant surprise for everybody.
It's not clear what the return on investment will be for a legislated rollout of broadband. In Canada, you are subsidizing broadband with your heavy taxes, as in Korea. If you read the article, broadband providers are hard-up for profits despite the high population density. How are Canadian providers doing?
You can make the chicken-egg argument that e-tailing, etc. would be more effective if more people had broadband. However, in my estimation, the issue was a lack of demand from those who had access, not a lack of access per se.
I think such an argument would be proven right for the _next_ generation, when today's kids become tomorrow's broadband consumers. But, by then, Korea-style broadband might be prevalent w/o gov't intervention, and the attendant inefficiencies.
If the country as a whole is moving up the foodchain, you need to move with it. If the software industry is outsourcing everything to India, be the person who decides what gets outsourced where.
Yeah, I know that there are complications due to gender, body type, genetics, etc. but if you are so fat that an unbalanced, low-cal diet is good for you, then a little exercise can't hurt.
A lot of exercise may not hurt, either.
As an American who grew up near the nation's capital, I hold the opposite view. The Concorde is cutting edge military technology misappropriated to a non-existent commercial market. On the other hand, the SR-71 is beautiful because it is purpose-built and every aspect of its design shouts out this purpose. The Blackbird's sole aim is reconnaissance, accomplished by flying higher and faster than anything coming from the ground.
Satellites obviate the SR-71, but for ~20 years it was an archangel.
... "mass piracy" is the object of utility, not the object of a transitive "threatens."
This is why Latin is superior for mechanistic parsing, because it would be the "ablative of utility" (or maybe the dative). These would be different cases than the accusative, which is usually the case for objects of transitive verbs in the active voice in the dead language.
Nevertheless, English flows well in speech and written text more or less, and has become the lingua franca for the Internet Age. Interesting stuff.
Yeah, this post is offtopic, but so is the parent. And I didn't find it funny, either.
Can it run uphill? Small internal combustion motors have very slim power bands, and I didn't see anything about gear shifting, nor a torque assist.
Will it last? I'm no mechanical engineer, but spur gears and an overrunning clutch do not sound like overly robust components.
All in all, I wonder if a moped is a better buy.
When I think of a planet, I think of a fluid (hot non-plasma gas, liquid metal) held together by self-gravity. Contrast with an asteroid, which is a rigid object (i.e., a rock). Furthermore, a planet must orbit around the sun, and not another body; if it does, it should be called a moon.
...
My $0.02
Isn't that application dependent?
What matters is that a distro a) does what it promises and b) is interoperable with other distros.
Requirement (b) is already handled by tarballs for most distros, and also in some by the low overhead in creating packages for them (e.g., Gentoo).
Requirement (a) is really what separates the distros. We'll see how Yoper fares. If two distros occupy the same niche, the better engineered one will win (well, hype not withstanding
Isn't that application dependent?
What matters is that a distro a) does what it promises and b) is interoperable with other distros.
Requirement (b) is already handled by tarballs for most distros, and also in some by the low overhead in creating packages for them (e.g., Gentoo).
Requirement (a) is really what separates the distros.
All things considered, I hope PDL, Python + extensions or Octave prove to be viable alternatives. As another insightful poster wrote, inertia/legacy is the biggest factor in deciding what to use. Some people use C/Fortran because it's the lowest common denominator, but the cure is worse than the disease. Good design, internally and in the user interface, is the way to go.
Now, when will there be a viable free alternative to Mathematia? Sorry, Maxima just doesn't cut it.
http://www.gentoo.org/dyn/pkgs/app-sci/octave.xml
Are there any viable open-source solutions to either Mathematica or IDL?
Does anybody know if MAS (or alternatively jack) can handle a driver for 3D audio rendering, not unlike DRI+Mesa in XF86?
CAVEAT: I'm of India descent, but I don't live in India, and the last time I was there was 1996.
....
I'd say most of the people who can read and write, and certainly those wealthy enough to have access to a computer, know English.
Local language support (Hindi is one, don't forget the other major languages like Bengali, Gujarati, Telugu, etc.) seems more appropriate when Linux usage extends beyond gov't/academia to home and commercial situations.
I skeptical that social change (i.e., adoption of computers/internet across the population) can be effected by supply-side pressure when there are such high barriers to adoption
What I put on the post to which you replied are guidelines, not dogma (hence my wistful surprise when I saw it get modded as flamebait!). I certainly indulge in cheese and crackers and sweets every now and then, and I don't limit myself vis-a-vis my nutrional regimen when I go out on engagements on Friday night or over the weekend. I find that my diet is pretty easy to stick to since a) it tastes good with a little planning and b) I feel better eating healthfully. I love a 3-inch think filet mignon, but tops once/month. When I get a hankering for red meat, I stick to lean cuts or lean ground ...
Happiness is the ultimate goal, no? I think health officials try to hard, and turn off people. I also think people try to hard, and turn off themselves. Very few people find true satisfaction in torturing themselves in a game of diminishing returns. I think I've found a comfortable critical point, as have you.
My own results have been mixed. I got pretty lean late last year when I had time to do things right, and my strength and endurance were quite good, but I didn't gain as much muscle mass as I wanted. I was probably overtraining, lifting four days a week an hour at a time, all out.
This dude is hardcore -- he's probably the top male fitness model out there right now. The only modification I've made is that I lift more and play basketball and do less cardio, and try to eat big after a workout to replenish my muscles.
What's worked for Slashdotters?