having spent quite a bit of the past week watching the slow deterioration of a channel i actually liked (well, substantially more than the rest of the crap we get) i've come to the conclusion that comcast wants to get rid of g4 and techtv both, and they're doing it by cutting off any pieces of techtv worth salvaging and then poisoning what's left with g4.
these numbers aren't surprising. seti is a volunteer project. with something like 85% of pcs in the world running windows (correct my figures if i'm wrong) you'd expect to see windows results dominating. as a side note, i'm not sure that seti is the best example of a/clustering/ application. as i recall, it's a grid application -- a different computation model.
i guess gates, etc. are suffering major tech envy over the fact that windows is still pretty much laughed at when it comes to serious computing. all the csi (computational sciences and informatics) labs at my university run linux now (they used to be indy workstations, now they're beefy dell boxen) and except for the professors' personal machines and the office machines, every single machine in the cs department runs some kind of unix.
ignoring the fact that the cs department has several important people who have a healthy hatred of microsoft, i don't think they'd every buy a windows cluster because (surprise, surprise) none of their software would work on it.
it sucks so badly that i couldn't find a better way to express just how abhorrent the experience of dealing with auto* is. i was going to say something about how legions of puppies and kittens and other cute little furry animals die torturous deaths at the hands of commie nazi pedophile terrorist hippies. urm. yes. you see the problems i was having;-P
not entirely sure why this got marked as a troll, but i guess that's what the meta-mod system is for. two things: firstly, it's available under some crazy dual license scheme which allows you the option of accepting either the QPL or the GPL (see here). secondly, i don't think that qmake comes even close to covering the ground auto(conf|make) does. i guess the closest relation is to automake, except crippled. as far as i know, it doesn't have any of the detection / configuration stuff that autoconf does. of course, it's been a while since i last used it, so that may have changed in newer versions.
i wish there was a way to moderate articles up, because you've hit on one of my major (*major*) pet psychotic hatreds regarding developing software. auto(conf|make) sucks badly. it's bearable if you're developing from scratch (not depending on other libraries) or require that your bundled versions of libraries be used. but when your software depends on, say, 123098123871237 other packages (i.e. you're writing for gnome or kde), you're boned.
unfortunately, there are no reasonable replacements that i know of, which is probably a testament to the nastiness inherent in solving this problem. a pity, really -- auto(conf|make) and company are a really good idea (in theory). unfortunately, there seems to be some really bad crack smoke involved in designing these tools. first (and probably foremost) in my mind is why isn't there a database of some sort which would at least allow the option of keeping track of which versions of what applications have been configured how and installed where.
it may be overkill for you, but i've used the handyboard (www.handyboard.com) for robotics stuff. it uses a c-esque language called interactive c, which is available gratis for linux, windows, and os x. it's a pretty decent (not great) environment. and if that fails, you can just program it in assembly (motorolla).
i'd be careful -- depends on the number of users. xml rpc / soap is hideously inefficient when it comes down to it. you're either wasting moderate amounts of time and huge amounts of bandwidth marshalling / unmarshalling requests or huge amounts of time and moderate amounts of bandwidth if you include compression in that process. if there aren't a lot of users, well who cares? however, if this app is supposed to support a substantial number of people, you run into problems. xml is not a silver bullet.
the answer depends: is your app a product? if so, it should probably support whatever rdbmses that your potential clients use. if it's a custom app for a client, do they use more than one rdbms? these are the only two cases in which it makes sense to spend substantial effort on isolating your database layer. otherwise, take the stored procedures and run -- this is exactly what they were designed for. there are a couple other advantages (other than performance) which may not have been mentioned:
1) abstraction. yes, abstracting the data layer at the database means accepting that you're tied to that database, which isn't a bad thing: i grew up with oracle and still swear by it. as a result, you only write one db layer api, not one per supported dbms per target business logic language.
2) depending on which rdbms and how you set up the user access, you can also probably push a large chunk of the connection pooling onto the db engine rather than having to manage it yourself.
3) access control. microsoft mentioned this several times at their devdays convention. i was surprised that it was news to them (hint: fewer doors means fewer ways in). regardless, it's a good idea: i've always seen allowing execution of arbitrary sql statements by joe random as an invitation for bad things.
4) one place to go to for developer support requests.
Probability and Statistics: I can't begin to put into words how useful this stuff is. As it happens, I use this stuff almost daily (doing probabalistic inference in computer vision). It also shows up in a lot of network classes (total reliability of the network, expected queue sizes and deviance, expected throughput and deviance, etc.) If you've understood this well, you'll laugh when you hear statistics like 40% of employee absenteeism happens on Mondays and Fridays while other people panick. You'll also not just to question the numbers you're given, but more importantly how to question them.
I get this "when are we ever going to use this stuff" schpiel from my lab students on a regular basis (I'm the lab T.A. for sections of CS1 and CS2). My usual response goes along these lines:
Physics: Calculus and Physics grew up together -- physics often having the problems which drove calculus. Understanding physics (or at least where the problems came from) tends to give enormous insight into how to actually solve problems. Physics also helps teach the basic science underneath Computer Science... experimental method, etc. Chemistry doesn't have the same deep developmental link with Calculus, but it does encourage the scientific mind.
Calc 1: This is the foundation of understanding how badly you're going to get screwed when you start haggling over how you're going to pay for a car / house / etc. It's also the foundation of Calc 2 and 3.
Calc 2: There's a huge class of problems which simply don't have a closed form solution or whose closed form solution. Things like Riemann sums, Taylor / MacLaurin series, etc. are unbelievably useful and crop up in the most unexpected places.
Calc 3: Now you can generalize all those things so that they work for functions of more than one variable. You start to appreciate why linear algebra is useful. This is where 90%+ of the math you need to do game physics and a large chunk of computer graphics originates.
Differential Equations: for most people, why we care about eigenvectors and eigenvalues. ECEs and EEs tend to care a lot more about this class. I liked it because I felt like my calculus was complete.
Discrete Math 1 + 2: (most people don't have to take Discrete 2) a large smorgasbord of things close to a computer scientists' heart: the combinatorics to figure out just how big a problem can be, the graph theory to understand data structures (not to mention how routing gets done on the internet), and a bunch of notation.
Fundamentals of Math: the math department's "how to do a proof" class. For us, this meant constructing the reals and rigorously proving all the things we usually take for granted about the reals, then extending this to rigorously develop a chunk of differential calculus.
Linear Algebra: I ended up taking two classes in this. The lower level class is pretty much a matrix cookbook / number crunching class. This was good enough for most computer graphics. I ended up enjoying the higher level class a lot because the proofs were really satisfying.
Numerical Analsysis: I can't stress how much I hated this class, but it was probably one of the most useful math classes I've taken. We ended up coding almost all the approximation techniques discussed in previous classes, and analysed their error bounds, with specific attention to where things break. We also touched on a lot of minimization techniques used in AI.
LaTeX comes with a whole suite of useful goodies, but there are some other really useful utilities you'll probably want to figure out. first is ispell (or aspell, haven't tried it), which is an interactive spell checker which (with the right cmdline switches) groks TeX (and thus LaTeX). next up is make, once your sources become fairly complicated (which for me means n >= 1 files usuall), a makefile becomes a real friend -- this way all your indices, glossaries, etc. are automagically regenerated as necessary, bibtex gets rerun as necesssary, etc. and (pdf)latex gets run until all crossreferences are resolved, if you have the right magic in your makefile.
what else... oh yeah, a couple word of advice: i'm a big fan of the amsmath and amssym packages (so math actually looks the way you expect it to). hyperref is nice if you want live links in your docs (so bibliography citations are linked to the bibliography entries, for example). i believe hyperref also lets you put in urls. there's a little weirdness in getting LaTeX to actually use an 8.5" x 11" page with 1" margins (it's fairly non-obvious). drop me a message and i'll show you the preamble that fixes this.
if you're going to be spending a lot of time writing up algorithms, a package like alg (or newalg) is pretty nice. i don't remember the specifics of its usage off hand, but if you check your handy dandy local ctan mirror (http://ctan.org), they'll have docs (+ sources) for all these packages and a ton more. there is a package which will even allow you to include C/C++/Java/Pascal/etc. code into your docs and pretty print that too (again, i forget which package, but i can check for you).
my weapon of choice for this stuff is emacs (yes, yes, insert flame war about how superior ${EDITOR} is to emacs, etc. etc. etc. *yawn*). this is partly out of inertia (i started with pico, was taught just enough emacs to accomplish a project as an intern and have sworn by it since), but mostly because emacs and i understand each other well. when handling LaTeX docs, emacs just does the Right Thing (tm), at least for the way i work. hit tab, and you'll find the current line indented to where it should be (magic!). lines break and wrap around (almost always) to where i think they should go. the only exception to this is when formatting the fairly hairy equations that i use (for some reason computer vision people do their damndest to indimidate with their notation).
i used _LaTeX for Linux_ by Bernice Sacks Lipkin (http://www.amazon.com/exec/obidos/tg/detail/-/038 7987088/qid=1083015560/sr=8-1/ref=sr_8_xs_ap_i1_xg l14/103-9939465-0827818?v=glance&s=books&n=507846) and it's a good cookbook-y kind of approach to things. i'm now fluent enough in LaTeX that when i do need to look something up, google does a good job of finding what i'm looking for.
btw, i started using LaTeX because )(*@#&* word wouldn't put pictures where i told it to/and/ properly crossreference captions. since then i'm the only kid who ever turns in anything that's LaTeX'd in any of my classes.
i doubt it, for for one major reason: there are entire classes of advantages and disadvantages which simply won't translate. how does the computer reward you for playing someone who's deluded into thinking that s/he is really zeus? (he he, that was a *fun* character) how do you handle the mental and social advantages / disadvantages / skills? for example, you're a historian from the future and you're doing some field research back in your favorite era. like the absent minded academic you are, you forgot to bring the right power converter to let your time machine work using the existing power supply, so you end up stranded with only one-way communications (if that!) to the future. translating this scenario into something that a computer could handle would be a daunting task (this is, of course, a fairly common occurance in the gurps universe[s] -- 50-100 points unusual background, -10 absent minded, etc.) how do you handle things like phobias, addictions, delusions, and so on?
if you look at the languages from whence java sprang, you'll see that while a lot of the semantics came from lisp (not case sensitive, easy to get lost in a sea of parentheses), the syntax is almost entirely based on that of C/C++/Object C (fewer parentheses, case sensitive). people coming from those languages (and there are a lot of them) are used to having case sensitivity and are likely to start bitching long and loud if you ever take it away from them.
short answer: any postscript printer should work just fine. longer answer: i've had my lexmark optra e 310 for nearly five years now, printing several hundred pages per month and i've had to change the toner exactly once. bonus: toner and imaging drum are all in one package, so it still prints as beautifully today as it did when i first bought it. yes, i know we all hate lexmark, but at least take a look at the optra e 220 or e 321.
how do you measure the degree to which the integrity of the data is preserved? psnr? mse? some other metric? i'm curious because jpeg2k is radically different from jpeg. and what in geospatial imaging qualifies as important data? reason i ask is that at least as far as computer vision goes, we lack the knowledge necessary to extract important image features other than texture, edgels, and "corners". in fact, there's often so much redundant information in an image that you pretty much have to throw away 80-95% of the image before you can start to understand what's going on. do the multispectral images you gis guys use have similar levels of redundancy? i'm curious to know -- all the people i know either do medical imagery (i.e. require lossless compression of huge images), document imaging (yay for jbig) or vision stuff, so i'm curious to know what rules you gis imaging folk play by;-)
Slashdot Mirror
having spent quite a bit of the past week watching the slow deterioration of a channel i actually liked (well, substantially more than the rest of the crap we get) i've come to the conclusion that comcast wants to get rid of g4 and techtv both, and they're doing it by cutting off any pieces of techtv worth salvaging and then poisoning what's left with g4.
these numbers aren't surprising. seti is a volunteer project. with something like 85% of pcs in the world running windows (correct my figures if i'm wrong) you'd expect to see windows results dominating. as a side note, i'm not sure that seti is the best example of a /clustering/ application. as i recall, it's a grid application -- a different computation model.
i guess gates, etc. are suffering major tech envy over the fact that windows is still pretty much laughed at when it comes to serious computing. all the csi (computational sciences and informatics) labs at my university run linux now (they used to be indy workstations, now they're beefy dell boxen) and except for the professors' personal machines and the office machines, every single machine in the cs department runs some kind of unix.
ignoring the fact that the cs department has several important people who have a healthy hatred of microsoft, i don't think they'd every buy a windows cluster because (surprise, surprise) none of their software would work on it.
i think billy & co finally figured out how to get big enough iron for longhorn >D
-$1,000,000,000,000,000,000,000,000,000,000.00 or -(bill gates' net worth), whichever is less.
it sucks so badly that i couldn't find a better way to express just how abhorrent the experience of dealing with auto* is. i was going to say something about how legions of puppies and kittens and other cute little furry animals die torturous deaths at the hands of commie nazi pedophile terrorist hippies. urm. yes. you see the problems i was having ;-P
not entirely sure why this got marked as a troll, but i guess that's what the meta-mod system is for. two things: firstly, it's available under some crazy dual license scheme which allows you the option of accepting either the QPL or the GPL (see here). secondly, i don't think that qmake comes even close to covering the ground auto(conf|make) does. i guess the closest relation is to automake, except crippled. as far as i know, it doesn't have any of the detection / configuration stuff that autoconf does. of course, it's been a while since i last used it, so that may have changed in newer versions.
i wish there was a way to moderate articles up, because you've hit on one of my major (*major*) pet psychotic hatreds regarding developing software. auto(conf|make) sucks badly. it's bearable if you're developing from scratch (not depending on other libraries) or require that your bundled versions of libraries be used. but when your software depends on, say, 123098123871237 other packages (i.e. you're writing for gnome or kde), you're boned.
unfortunately, there are no reasonable replacements that i know of, which is probably a testament to the nastiness inherent in solving this problem. a pity, really -- auto(conf|make) and company are a really good idea (in theory). unfortunately, there seems to be some really bad crack smoke involved in designing these tools. first (and probably foremost) in my mind is why isn't there a database of some sort which would at least allow the option of keeping track of which versions of what applications have been configured how and installed where.
it may be overkill for you, but i've used the handyboard (www.handyboard.com) for robotics stuff. it uses a c-esque language called interactive c, which is available gratis for linux, windows, and os x. it's a pretty decent (not great) environment. and if that fails, you can just program it in assembly (motorolla).
... when will i be able to use it to cool my system so i can overclock another 100MHz out of it and make winamp *that* much faster?
i'd be careful -- depends on the number of users. xml rpc / soap is hideously inefficient when it comes down to it. you're either wasting moderate amounts of time and huge amounts of bandwidth marshalling / unmarshalling requests or huge amounts of time and moderate amounts of bandwidth if you include compression in that process. if there aren't a lot of users, well who cares? however, if this app is supposed to support a substantial number of people, you run into problems. xml is not a silver bullet.
the answer depends: is your app a product? if so, it should probably support whatever rdbmses that your potential clients use. if it's a custom app for a client, do they use more than one rdbms? these are the only two cases in which it makes sense to spend substantial effort on isolating your database layer. otherwise, take the stored procedures and run -- this is exactly what they were designed for. there are a couple other advantages (other than performance) which may not have been mentioned:
1) abstraction. yes, abstracting the data layer at the database means accepting that you're tied to that database, which isn't a bad thing: i grew up with oracle and still swear by it. as a result, you only write one db layer api, not one per supported dbms per target business logic language.
2) depending on which rdbms and how you set up the user access, you can also probably push a large chunk of the connection pooling onto the db engine rather than having to manage it yourself.
3) access control. microsoft mentioned this several times at their devdays convention. i was surprised that it was news to them (hint: fewer doors means fewer ways in). regardless, it's a good idea: i've always seen allowing execution of arbitrary sql statements by joe random as an invitation for bad things.
4) one place to go to for developer support requests.
hope this helps.
D'oh, I forgot:
I get this "when are we ever going to use this stuff" schpiel from my lab students on a regular basis (I'm the lab T.A. for sections of CS1 and CS2). My usual response goes along these lines:
you mean his net worth decreased from 46.6$ billion to... 46.6$ billion. hrm...
the more obvious question is: why pay for dcl when there's libz and libbz2 available open /and/ free?
i will have to try that -- i have a substantially nastier setup:
d dsidemargin}{1in}e tlength{\voffset}{-0.5in}{ 6.5in}
\setlength{\evensidemargin}{1in}
\setlength{\o
\setlength{\hoffset}{-1in}
\s
\setlength{\textwidth}
\textheight=9.0in
LaTeX comes with a whole suite of useful goodies, but there are some other really useful utilities you'll probably want to figure out. first is ispell (or aspell, haven't tried it), which is an interactive spell checker which (with the right cmdline switches) groks TeX (and thus LaTeX). next up is make, once your sources become fairly complicated (which for me means n >= 1 files usuall), a makefile becomes a real friend -- this way all your indices, glossaries, etc. are automagically regenerated as necessary, bibtex gets rerun as necesssary, etc. and (pdf)latex gets run until all crossreferences are resolved, if you have the right magic in your makefile.
what else... oh yeah, a couple word of advice: i'm a big fan of the amsmath and amssym packages (so math actually looks the way you expect it to). hyperref is nice if you want live links in your docs (so bibliography citations are linked to the bibliography entries, for example). i believe hyperref also lets you put in urls. there's a little weirdness in getting LaTeX to actually use an 8.5" x 11" page with 1" margins (it's fairly non-obvious). drop me a message and i'll show you the preamble that fixes this.
if you're going to be spending a lot of time writing up algorithms, a package like alg (or newalg) is pretty nice. i don't remember the specifics of its usage off hand, but if you check your handy dandy local ctan mirror (http://ctan.org), they'll have docs (+ sources) for all these packages and a ton more. there is a package which will even allow you to include C/C++/Java/Pascal/etc. code into your docs and pretty print that too (again, i forget which package, but i can check for you).
hope this helps.
my weapon of choice for this stuff is emacs (yes, yes, insert flame war about how superior ${EDITOR} is to emacs, etc. etc. etc. *yawn*). this is partly out of inertia (i started with pico, was taught just enough emacs to accomplish a project as an intern and have sworn by it since), but mostly because emacs and i understand each other well. when handling LaTeX docs, emacs just does the Right Thing (tm), at least for the way i work. hit tab, and you'll find the current line indented to where it should be (magic!). lines break and wrap around (almost always) to where i think they should go. the only exception to this is when formatting the fairly hairy equations that i use (for some reason computer vision people do their damndest to indimidate with their notation).
i used _LaTeX for Linux_ by Bernice Sacks Lipkin (http://www.amazon.com/exec/obidos/tg/detail/-/038 7987088/qid=1083015560/sr=8-1/ref=sr_8_xs_ap_i1_xg l14/103-9939465-0827818?v=glance&s=books&n=507846) and it's a good cookbook-y kind of approach to things. i'm now fluent enough in LaTeX that when i do need to look something up, google does a good job of finding what i'm looking for.
/and/ properly crossreference captions. since then i'm the only kid who ever turns in anything that's LaTeX'd in any of my classes.
btw, i started using LaTeX because )(*@#&* word wouldn't put pictures where i told it to
i doubt it, for for one major reason: there are entire classes of advantages and disadvantages which simply won't translate. how does the computer reward you for playing someone who's deluded into thinking that s/he is really zeus? (he he, that was a *fun* character) how do you handle the mental and social advantages / disadvantages / skills? for example, you're a historian from the future and you're doing some field research back in your favorite era. like the absent minded academic you are, you forgot to bring the right power converter to let your time machine work using the existing power supply, so you end up stranded with only one-way communications (if that!) to the future. translating this scenario into something that a computer could handle would be a daunting task (this is, of course, a fairly common occurance in the gurps universe[s] -- 50-100 points unusual background, -10 absent minded, etc.) how do you handle things like phobias, addictions, delusions, and so on?
if you look at the languages from whence java sprang, you'll see that while a lot of the semantics came from lisp (not case sensitive, easy to get lost in a sea of parentheses), the syntax is almost entirely based on that of C/C++/Object C (fewer parentheses, case sensitive). people coming from those languages (and there are a lot of them) are used to having case sensitivity and are likely to start bitching long and loud if you ever take it away from them.
i think i've had one paper jam ever and that was when i fed the machine too much paper :-/
short answer: any postscript printer should work just fine. longer answer: i've had my lexmark optra e 310 for nearly five years now, printing several hundred pages per month and i've had to change the toner exactly once. bonus: toner and imaging drum are all in one package, so it still prints as beautifully today as it did when i first bought it. yes, i know we all hate lexmark, but at least take a look at the optra e 220 or e 321.
how do you measure the degree to which the integrity of the data is preserved? psnr? mse? some other metric? i'm curious because jpeg2k is radically different from jpeg. and what in geospatial imaging qualifies as important data? reason i ask is that at least as far as computer vision goes, we lack the knowledge necessary to extract important image features other than texture, edgels, and "corners". in fact, there's often so much redundant information in an image that you pretty much have to throw away 80-95% of the image before you can start to understand what's going on. do the multispectral images you gis guys use have similar levels of redundancy? i'm curious to know -- all the people i know either do medical imagery (i.e. require lossless compression of huge images), document imaging (yay for jbig) or vision stuff, so i'm curious to know what rules you gis imaging folk play by ;-)