Why do you miss apt-get? I've never been a Debian user, but I've heard arguments from people at my local linux user group, and they're all obscenely biased.
The process of selection is done by assigning reviews to major professors in the field who are not submitting to the conference.
Bless, how naive. I'm a researcher in a red-brick university in England, and I know that to not always hold true. I've known cases where a submitter has received other peoples work for evaluation, along with an advice sheet recommending that they are forgiving since not enough papers had been put forward. They are also not always professors by any stretch of the imagination, I know people who have assessed submissions that lack even a PhD.
Reviewers are also required to find ALL spelling and grammer mistakes
Hoho. If that were honestly true then you wouldn't find any spelling and grammar mistakes in published papers. That's just not true.
Papers can and are quickly manufactured. You're clearly not reading around enough if you honestly believe that not to be so. In my current area there's all sorts of gaseous crap that gets written about Grid computing. Pick an upcoming area, research lots of long words, string them together, suggesting that it will be possible to develop a synergistic solution with another hot topic, and you've got a paper. Oops, forgot the graphs.
Just to defend my little island, you can get them no problem in Great Britain in a couple of different types.
And I'd reckon we're slightly less confused about the state of the rest of the world. One of the benefits of our previous colonialist policies is that we have much closer ties with these third world countries than many other nations.
But they're in the business of making money, not doing good for the people. If they can make more money out of it by not GPLing it, then they should keep it.
Also, how long did it take to develop the Quake III engine? How many people? They're free to recoup their costs as they like.
You're making the assumption that they are in fact going to come together, and that's a non-trivial assumption. Why can't the objects in the system simply carry on moving further and further away, without ever returning. You are missing the possibility of the gravitational forces reducing in effect due to distance at such a rate that the objects do not return.
Whether or not there is this 'axis' you talk of, depends on the amount of mass present in the universe. Below a certain threshold, they'll go forever, above and they'll all come back in a big crunch.
Simply moving the objects further away wouldn't increase their kinetic energy. That would only be true if the speed at which they were travelling at was still increasing. Is it? I can't see from my office.
As I understood it more *money* went into fortran compilers. Since people were really concerned about speed, compilers could cost an arm and a leg and still get bought. If optimising fortran is so easy, why's gnu f77 such a pile of crap?
This shows absolutely no understanding of the language. Sit down if you're not in the field.
I used to teach "Practical Parallel Programminh" at the Univesity of Leeds and this is just crap. Fortran is typically used with OpenMP / MPI to do parallel programming. Older freaks might use PVM. They're all available for C/C++.
And it's not that I'm no longer in the field, I currently work on Grid/Globus applications.
Fortran is no more safe or fast to program in I'd argue it's a less safe myself. The performance difference between an optimal fortran program and an optimal C program I'd argue is nearly nil. Show me different, and explain why. Go on, try it.
Deleting deprecated methods is fair enough. They're a support nightmare if they get any more out of hand (since all development on Java must maintain the functionality of deprecated methods). You've got to do it at somepoint, and either you slowly throw them out, or do it in one swift cull.
The naming of methods/classes was wrong. When you learn java you realise it straight away. That would be a very minor pain to bear in terms of fixing old code.
From the beginning I said Java hadn't gone far enough and made the mistake of having primitive types. Don't tell me you thing Integer et al is good.
You appear to completely misunderstand the move to XML. What's the significance to servlets? The current serialisation format is bloaty and slow, not to mention the problems of versioning (which can be made easier).
I think you're making light of his comments, they certainly echo (and exceed) the suggestions I've made within my research group.
I don't think it's quite like you're saying, since there isn't the automatic copying. This is memory that *can* be addressed by your graphics card AFAIK. Intel claim it automatically limits itself to the value specified, or available RAM, whichever is smaller.
Benchmarks and matroxusers.com convinced me to use 256Mb with the G200. The stability issue with my machine is a whole different matter... I've yet to find something that will explain why an AGP Aperture setting would make a machine unstable (I couldn't do a winXP install, DRI caused lockups...)
The card in question was a Hercules 3d Prophet 2 MX, on an Epox 8K3A KT333 Motherboard.
I'm guessing that it's the AGP Aperture size. It kinda depends on a the graphics card, but I'd definitely set it higher. 256Mb was the best for my Matrox G200 (8Mb graphics card, 224 Mb System), and 128Mb is the only stable size for my GeForce2 (32Mb gfx, 512Mb System).
I'll second that. I have reasonably recent versions of gvim installed on WinME + WinXP boxen with no problems whatsoever. Once I fix the defaults back to more like the standard unix defaults it works just like I want it, and I've certainly *never* had it crash.
Looks like he's using an odd version, or his windows install is so borked it can't run anything reliably, or he's got dodgy hardware.
RAID does NOT improve performance. It hurts it. Read some benchmarks if you don't believe.
You're joking right? Are you a RAID hater or what? Let's say you're using straight striping (Raid 0). That improves performance. Don't even argue it, since it just stripes, there's not even a checksum overhead. If you're using RAID 5 then there are more complex considerations, since there is far more than just writing data. But to say that it does not improve performance just shows that you're not a RAID user.
Also if you think that the difference between SCSI and IDE is CPU load, I also suggest you smell the coffee.
In many cases, they are the exact same drive with different electronics attached to them.
So we'll just ignore the 15k rpm SCSI units then shall we? Or how about just the 10k units? And we won't even get into looking at the 2 unit per bus limit on IDE will we. Or the lack of external connectors.
I've also got about 10 desktop computers flipped on their side, with 'server' written on them in crayon.
Can we just not go there? You clearly have no idea about running a proper high availability server setup.
If it can be down, just keep it backed up, and buy cheap.
I've always just done searches for "unplugged" and "live", taken wild guesses looking at the results and had a listen. That's given me some damn fine music that I never would have got near otherwise. Probably only bought 4-5 albums off the back of it, but that's out of at most 10 I bought in a year.
The price per unit has no bearing on how much you use. You're making a really poor point here. We're not talking about using less dollar value electricity, we're talking about generating less electricity.
This guy's not after a uni with good beer. I'm guessing he's assessing them from the role of an educator. As such here's my opinions coming from someone who actually teaches this stuff, AND takes teaching methodolgies seriously:
1. How many different languages do you expose the undergraduates to? Why? Why those languages?
I want them to have a good understanding of why it's important for their graduates to be language neutral when they leave. I want them to have a good understanding of the advantages of teaching functional/procedural languages.
2. If they're teaching a loosely OO language like C++ I want to know, "Are you teaching objects first?". One of the current trends in England is to teach objects first in C++, despite the fact that it causes information overload in the early stages of teaching.
3. Do they integrate modules? One problem with many British universities is they've adopted a modular approach to courses to enable students to pick and choose what they would like to learn. The problem with this approach is that it does not make it possible to tie in modules, since the pre-requisites become too complicated. Their opinions on this problem would be of great interest.
4. At our university, some high level academics are of the opinion that programming is a old-school skill that will be replaced by IDEs + CASE tools. Stupid concept, but I'd like to try and catch people that think it. Their opinions on the role of tools such as Rational Rose and UML would be informative.
5. Whether or not they feel that psychology modules are important. Some try to wrap it up as Human Computer Interaction, but farming it out to the psych dept and getting thorough coverage of ethics in computing, and the psychology of colour etc makes for a very interesting and diverse syllabus.
6. Importance of teaching hardware + real time systems.
7. Teaching of proprietary APIs. Again to take my own dept as an example, we teach SQL Server specific SQL. Now what's the crack with that? Yet at the other end of the scale, we teach OpenGL/g2/pthreads/MPI.
8. Use of different operating systems. We're now a bit lightweight on that front since retiring our ageing IRIX machines, but we expose students to at least Linux + Windows 2000. SunOS is possible, IRIX slightly less so.
9. Teaching of parallel programming techniques, both theoretical (CCS + other formal logic) and practical MPI/threads. It's a great programming challenge that teaches skills that are relevant in traditional programming too.
Same with VMWare. You can run multiple instances, and either network them just with each other, or bridge them out so that they appear on your network with their own IPs.
It asks you a couple of questions during the OS setup and it worked out of the box for me.
That's such a simplistic attitude to take. Yes you can saturate your network assuming that you're pumping out full data rates, but that's not the point and never was. What about when you have 50 people (a *very* conservative number) accessing lots of small files from these disks. SCSI will kill you hands down due to better drive logic, and command queuing. You're trying to apply simplistic maths to a complex problem.
We have one set of machines with IDE drives and another set with Ultra160 drives. I'll tell you which I'm choosing to sit at, and it's no small performance difference either. And that's on a desktop. Linux is pretty good at opening 101 rc files and the like. Slow disks gives slow linux, and I know SCSI is a killer.
I think you're highly confused as to the role of the X server and that of the window manager.
Integration of everything into X isn't the answer you're looking for.
But if you think it is, fine, it doesn't require anything major, since the pluggable design of X allows you to add what you like.
I remember from a discussion on the design of X, the designers suggesting that even the basic primitives should be modules, but decided that a windowing system that was unable to draw a window without a plugin was a bad idea...;)
Slashdot Mirror
But what does that do? Term it for a non apt-get user.
That downloads the dependencies required for building the package from source? I'm not certain here.
I'm not sure urpmi supports anything source related yet, but I could be wrong.
Okay, carry on, what else...
jh
Why do you miss apt-get? I've never been a Debian user, but I've heard arguments from people at my local linux user group, and they're all obscenely biased.
What can apt-get do that urpmi can't?
Bless, how naive. I'm a researcher in a red-brick university in England, and I know that to not always hold true. I've known cases where a submitter has received other peoples work for evaluation, along with an advice sheet recommending that they are forgiving since not enough papers had been put forward. They are also not always professors by any stretch of the imagination, I know people who have assessed submissions that lack even a PhD.
Hoho. If that were honestly true then you wouldn't find any spelling and grammar mistakes in published papers. That's just not true.
Papers can and are quickly manufactured. You're clearly not reading around enough if you honestly believe that not to be so. In my current area there's all sorts of gaseous crap that gets written about Grid computing. Pick an upcoming area, research lots of long words, string them together, suggesting that it will be possible to develop a synergistic solution with another hot topic, and you've got a paper. Oops, forgot the graphs.
Sorry I've just become a cynic.
jh
Just to defend my little island, you can get them no problem in Great Britain in a couple of different types.
And I'd reckon we're slightly less confused about the state of the rest of the world. One of the benefits of our previous colonialist policies is that we have much closer ties with these third world countries than many other nations.
jh
Mod this crap down. The whole point is they don't want something ntp based. Sheesh.
Can I safely assume you're not a developer? If you are, how much to hire, you sound outlandishly good.
jh
But they're in the business of making money, not doing good for the people. If they can make more money out of it by not GPLing it, then they should keep it.
Also, how long did it take to develop the Quake III engine? How many people? They're free to recoup their costs as they like.
jh
Missing the point that they could force them to release the finished product through ID if they actually want to make any money off it.
That would become the revenue stream.
jh
Nope.
Study astrophysics theory and you'll find this is actually well within the realms of reality, and is a much debated point.
jh
You're making the assumption that they are in fact going to come together, and that's a non-trivial assumption. Why can't the objects in the system simply carry on moving further and further away, without ever returning. You are missing the possibility of the gravitational forces reducing in effect due to distance at such a rate that the objects do not return.
Whether or not there is this 'axis' you talk of, depends on the amount of mass present in the universe. Below a certain threshold, they'll go forever, above and they'll all come back in a big crunch.
jh
Simply moving the objects further away wouldn't increase their kinetic energy. That would only be true if the speed at which they were travelling at was still increasing. Is it? I can't see from my office.
jh
As I understood it more *money* went into fortran compilers. Since people were really concerned about speed, compilers could cost an arm and a leg and still get bought. If optimising fortran is so easy, why's gnu f77 such a pile of crap?
jh
This shows absolutely no understanding of the language. Sit down if you're not in the field.
I used to teach "Practical Parallel Programminh" at the Univesity of Leeds and this is just crap. Fortran is typically used with OpenMP / MPI to do parallel programming. Older freaks might use PVM. They're all available for C/C++.
And it's not that I'm no longer in the field, I currently work on Grid/Globus applications.
Fortran is no more safe or fast to program in I'd argue it's a less safe myself. The performance difference between an optimal fortran program and an optimal C program I'd argue is nearly nil. Show me different, and explain why. Go on, try it.
jh
Deleting deprecated methods is fair enough. They're a support nightmare if they get any more out of hand (since all development on Java must maintain the functionality of deprecated methods). You've got to do it at somepoint, and either you slowly throw them out, or do it in one swift cull.
The naming of methods/classes was wrong. When you learn java you realise it straight away. That would be a very minor pain to bear in terms of fixing old code.
From the beginning I said Java hadn't gone far enough and made the mistake of having primitive types. Don't tell me you thing Integer et al is good.
You appear to completely misunderstand the move to XML. What's the significance to servlets? The current serialisation format is bloaty and slow, not to mention the problems of versioning (which can be made easier).
I think you're making light of his comments, they certainly echo (and exceed) the suggestions I've made within my research group.
jh
I don't think it's quite like you're saying, since there isn't the automatic copying. This is memory that *can* be addressed by your graphics card AFAIK. Intel claim it automatically limits itself to the value specified, or available RAM, whichever is smaller.
Benchmarks and matroxusers.com convinced me to use 256Mb with the G200. The stability issue with my machine is a whole different matter... I've yet to find something that will explain why an AGP Aperture setting would make a machine unstable (I couldn't do a winXP install, DRI caused lockups...)
The card in question was a Hercules 3d Prophet 2 MX, on an Epox 8K3A KT333 Motherboard.
jh
I'm guessing that it's the AGP Aperture size. It kinda depends on a the graphics card, but I'd definitely set it higher. 256Mb was the best for my Matrox G200 (8Mb graphics card, 224 Mb System), and 128Mb is the only stable size for my GeForce2 (32Mb gfx, 512Mb System).
HTH,
jh
I'll second that. I have reasonably recent versions of gvim installed on WinME + WinXP boxen with no problems whatsoever. Once I fix the defaults back to more like the standard unix defaults it works just like I want it, and I've certainly *never* had it crash.
Looks like he's using an odd version, or his windows install is so borked it can't run anything reliably, or he's got dodgy hardware.
jh
I've always just done searches for "unplugged" and "live", taken wild guesses looking at the results and had a listen. That's given me some damn fine music that I never would have got near otherwise. Probably only bought 4-5 albums off the back of it, but that's out of at most 10 I bought in a year.
jh
The price per unit has no bearing on how much you use. You're making a really poor point here. We're not talking about using less dollar value electricity, we're talking about generating less electricity.
Sheesh.
No wonder you guys pulled out of Kyoto.
jh
This guy's not after a uni with good beer. I'm guessing he's assessing them from the role of an educator. As such here's my opinions coming from someone who actually teaches this stuff, AND takes teaching methodolgies seriously:
1. How many different languages do you expose the undergraduates to? Why? Why those languages?
I want them to have a good understanding of why it's important for their graduates to be language neutral when they leave. I want them to have a good understanding of the advantages of teaching functional/procedural languages.
2. If they're teaching a loosely OO language like C++ I want to know, "Are you teaching objects first?". One of the current trends in England is to teach objects first in C++, despite the fact that it causes information overload in the early stages of teaching.
3. Do they integrate modules? One problem with many British universities is they've adopted a modular approach to courses to enable students to pick and choose what they would like to learn. The problem with this approach is that it does not make it possible to tie in modules, since the pre-requisites become too complicated. Their opinions on this problem would be of great interest.
4. At our university, some high level academics are of the opinion that programming is a old-school skill that will be replaced by IDEs + CASE tools. Stupid concept, but I'd like to try and catch people that think it. Their opinions on the role of tools such as Rational Rose and UML would be informative.
5. Whether or not they feel that psychology modules are important. Some try to wrap it up as Human Computer Interaction, but farming it out to the psych dept and getting thorough coverage of ethics in computing, and the psychology of colour etc makes for a very interesting and diverse syllabus.
6. Importance of teaching hardware + real time systems.
7. Teaching of proprietary APIs. Again to take my own dept as an example, we teach SQL Server specific SQL. Now what's the crack with that? Yet at the other end of the scale, we teach OpenGL/g2/pthreads/MPI.
8. Use of different operating systems. We're now a bit lightweight on that front since retiring our ageing IRIX machines, but we expose students to at least Linux + Windows 2000. SunOS is possible, IRIX slightly less so.
9. Teaching of parallel programming techniques, both theoretical (CCS + other formal logic) and practical MPI/threads. It's a great programming challenge that teaches skills that are relevant in traditional programming too.
Anyhows, just my thoughts, hope they help,
jh
Same with VMWare. You can run multiple instances, and either network them just with each other, or bridge them out so that they appear on your network with their own IPs.
It asks you a couple of questions during the OS setup and it worked out of the box for me.
jh
READ THE FSCKING ARTICLE.
He tested both under x86 and using a G4 Cube.
This author deserves credit, so quit your whining, and clap. It's not the most in depth review possible, but it's a good short piece.
jh
That's such a simplistic attitude to take. Yes you can saturate your network assuming that you're pumping out full data rates, but that's not the point and never was. What about when you have 50 people (a *very* conservative number) accessing lots of small files from these disks. SCSI will kill you hands down due to better drive logic, and command queuing. You're trying to apply simplistic maths to a complex problem.
We have one set of machines with IDE drives and another set with Ultra160 drives. I'll tell you which I'm choosing to sit at, and it's no small performance difference either. And that's on a desktop. Linux is pretty good at opening 101 rc files and the like. Slow disks gives slow linux, and I know SCSI is a killer.
jh
I think you're highly confused as to the role of the X server and that of the window manager.
;)
Integration of everything into X isn't the answer you're looking for.
But if you think it is, fine, it doesn't require anything major, since the pluggable design of X allows you to add what you like.
I remember from a discussion on the design of X, the designers suggesting that even the basic primitives should be modules, but decided that a windowing system that was unable to draw a window without a plugin was a bad idea...
jh