One reason that there is a stigma attached to non-peer review and net based papers is that you can write any old garbage. I should know, my most cited paper (by a factor of 5) is in an electronic journal (It's not garbage, but it is informational rather than scientific peice).
Publication lists are often used by funding boards to asses the credibility of a proposal from an author. While less than ideal, this is necessary because it would be impractical for every funding board to review every aspect of every proposers work - they have neither the time, or the breadth of expertise. If they included non-peer-review publications, then there would be some people who like writing churning out heaps of garbage papers. (I've reviewed some which are clearly rubbish. I've also asked for the editorial board to pick another reviewer when I disagree with the author concerned).
At the same time it would be wrong to deny that there are a few fields in which a viewpoint has acheived 'monopoly status', and alternate views cannot be expressed, although it is uncommon for all the journals in a field to be so dominated unless the field is very narrow. Experience shows that the situation usually resolves itself as the current generation of reviewers retire, although that is frustrating for individual authors at the time.
Given the balance between not abusing the nations tax $ on project with no scientific merit, and not holding up the progress of science, I think the balence in the current system is about right, even if it goes wrong in a few specific cases. If peer-reviewed electronic publishing can reduce the number of bad cases, I'm all for it.
I agree that peer-review is vital to filter out crackpots and commercial propaganda. But rather than the traditional editorial board approach, why not slashdot style moderation?
Give each reader and each paper a credibitlty score. The credibility of the reader is based on the average credilibity of each paper they have submitted. Each reader can give a credibility score to each paper they read. The score for the paper will be the submitted scores, weighted by the credibility of each reviewer.
Now it is easy to make foulups in a paper the first time round, so an initial submission could be made to an editorial area. People could add comments slashdot style, and the authors could use these to revise the paper for final submission.
When I started reading this, I hated it. But now I love it! If someone launches a journal in my field this way, I'll gladly submit a paper. Of course, there is no real need to have individual journals, if the database can be searched flexibly.
I agree with you in prinicle, but not in practice.
The GUI is important because these days anything with a GUI is comfortingly familiar, and anything without is dangerous and frightening. But, the problem of not adding bloat before the bulk of the install is an important one.
I'm not sure perl is an issue, since I think a perl runtime is quite small and available for many platforms. But X, gtk and so on are certainly an issue.
To me the ideal approach seem to be a fake GUI which dresses up the CLI install - just like Stormix are doing. You can use a 16 colour VGA driver, which is small and will work on almost anything, and still draw good 'fake' gtk or qt widgets and window decorations. If it is done properly (and it looks like Stormix is), then the CLI option can be run with the same code, and I guess the bloat for the CLI can be kept to a fraction of a floppy.
I'm not sure whether this is the approach adopted by either Caldera or Stormix, or whether they are starting up a basic x-server. But its the way I'd do it.
I thought at first that writing it in perl was a bad idea, because you need a perl runtime environement (or does perl compile native these days?).
But one big advantage of using perl is that the port to Win32 should be fairly easy. Assuming the disk access code for windows can be ripped from FIPS and put in a library, the trickiest part will be the GUI. Of course, if they used a cross platform toolkit (either wxWindows or QT) then that would become a non-issue.
It is on X, but you can't tell X by looking. X is a protocol for talking to graphical terminals, just as VT100 was a protocol for talking to text terminals. This gets confusing with workstations, becuase the terminal and the host are the same computer.
Systems running X can look completely different, even Linux systems, depending on whether they use 75dpi or 100 dpi fonts, which window manager is running, which widget set and application uses, which theme the window manager and widget set is using (where applicable).
Things are further confused by look-alike themes. In some cases a Mac, or W95 theme running on X can be indistiguishable from the real thing.
There's very little point in it unless you have an interesting problem, and a heap of data.
But if you do, then there is suddenly a great deal of point. Unfortunately I don't think the machine they are proposing is very useful - it certainly wouldn't run any of my jobs, which will require a lot of work to migrate from vector supercomputers to any sort of parallel machine.
Even for jobs which can be massively parallelised, there is always the problem of compiling the results. Even for very simple tasks like searching for primes or crypto keys this becomes a headache with so many machines. For anything more complex, it will require a supercomputer with multiple fibre channels just to combine the results.
Where they are going to get a couple of megawatts to run this thing. Also, enough airconditioning to get rid of a couple of megawatts of heat.
Clearly the networking of this many machines will be impractical for most mathematical problems, I guess this will only be good for problems which could already be tackled by d.net or seti.
I think you've missed the point of the 200MHz system bus. The main benefit is not a faster path to the main memory, but sustained fast access to multiple paths, thus A DMA can be running at the same time as CPU access without a big slowdown.
Further, initial tests show the RAMBUS memory which will be able to keep up with faster system bus looks as though it will actually run slower in most cases, because although the data rate is higher, the latency (the time taken to start a transfer) is much longer. Only very special applications will benefit, most will run slower.
Search www.theregister.co.uk for lengthy discussions of RAMBUS and its problems. PC133 looks like a better bet at the moment, but Intel has bet the house on RAMBUS.
Since a geosynch orbit has a radius of 42000 Km (22000 mi), and the Earth has a radius of 6400 Km (4000 mi), assuming zero axial tilt, a geosynch collector will be in sunlight for 22 hours 50 minutes each day.
Once you take into account the axial tilt, this will increase to very nearly 24 hours during the summer and winter or the ground station.
IIRC the reason microwave ovens work is that the microwaves are tuned to the energy of a bond in the water molecule. Thus microwaves heat up water. They don't heat up ice, and dry foods heat only slowly.
The problem is to get lots of energy from orbit to earth. Using mirrors and light would work, but light is absorbed by lots of things, and anything which absorbs lots of energy in any form is going to suffer.
So the challenge is to use a radiation which will go straight through almost anything that might stray into its path, but can still be picked up and converted by the receiving station. I guess the researchers have picked a band in the microwave spectrum which doesn't interact strongly with water or protein bonds - I don't know how easy this is. The receiver just needs an antenna designed for the particular band (this might be an array of simple rod-type dipoles, or a big horn/waveguide).
Of course (metal) airplanes will be a problem, but there are already plenty of no-fly-zones on the globe, I don't see a few more being a problem. (I assume the collector will be geostationary).
Well, you can use multiple network cards, but it makes the network run slower, not faster, hence the Mindcraft debacle. So I guess multiple NIC's aren't critical to a Beowulf.
Of course, the new threaded TCP/IP stack is now in the development version, so Linux should soon excell at benchmarks, and it might help cluster apps too.
The open source model is a close relative of the scientific development model, with code treated as research, and published for peer review and for the benefit of mankind.
And the problems are very similar. I've seen equivalent behaviour in the scientific community. Perhaps J was a little younger than many scientists - normally such disputes are played out in a manner more mature, although no less vicious.
Don't worry about it. Its human nature. The best we can do is remember to laugh at ourselves.
A fable-like tale, set in the Middle Ages, focuses on two outcasts, an atheist prince and a wolf child confined to a nunnery, who become pawns in a theological controversy. Booker Prize nominee. Reprint.
It reminded me of Name of the Rose, without being heavy.
I wasn't going to reply to this, because I really don't know enough literary theory to say whether 'Lord of the Rings' is good writing or not. It is certainly popular, but that is something different: Citizen Cane is great cinema, but not popular. 'The Sun' is the UK's most popular newspaper, but no-one claims it is good (not even many of the writers). It is enjoyable, but if it were not it would not be popular.
But what I can do is compare it with other books. Now I admit that if you compare LotR with a lot of other fantasy, it comes off quite well.
It is more difficult to compare across genres, but I think you can compare universal elements such as prose style and characterisation. And when I compare the style with some mainstream novels, it seems somehow lacking:
For an example of deep and moving characterisation, try 'Portrait of a Lady' by Henry James.
For an example of beautiful prose, try '100 years of Solitude' by Gabriel Garcia Marquez.
Iain Banks 'The Bridge' and 'Walking on Glass' paint vivid worlds. The latter has a gripping plot or three.
I don't think the literary style in LotR can seriously be compared with any of these.
I wouldn't recommend the first two as fun books to read, the first is depressing and the second is very slow, but at the same time, both are unforgetable. The two Banks novels I mention because the first is science fiction(ish), and the second has fantasy elements.
What is the point I am trying to make? I'm not really sure. But it is something like this:
There are books you read and remember for the rest of your life. And there are books you grab to read on a plane because it is fun and passes the time, but once you throw them away they're forgotten.
Going back to LotR after 15 years and a lot of reading, it feels as though it should be in the second group. But it isn't. The only reason I can give is that it was my first fantasy, just as 'Islands in the Sky' was my first SF, and both have a special place, not for any particular technical merit, but because they were first.
But then again, I don't know anything about writing. All I can say is that I've read a bit of everything.
It's a funny one this. I almost never swear, I'm put off by swearing (in real life), I'm a Christian. I'm frequently bored by stand up comics who think they can get a laugh by adding a few obscenities to an otherwise dull routine, and depressed to find that often they can.
South Park was the funniest movie I have ever seen.
I guess the issue might be that South Park pokes fun at the Moral Majority. If you're not part of it, and don't take yourself too seriously, then its funny.
Here's another odd thing: Sabrina, Buffy and Daria all feature attractive heroines who feel rejected by the fashionable set. Are they pandering to geek fantasies, or the inadaquacy/rejection thing is far more widespread, with personal significance to the majority of the population? I suspect the latter.
If they were being paid, maybe so. But since most of the developers work for the satisfaction of it, I don't think we really have much say in the matter.
And of course, the competition leads to improvement and innovation.
Before the rise of the railways, it was normal for cities in Britain, and most probably everywhere else as well, to use their own local time. Thus, in traveling a few hundred miles east or west, you would have to reset your watch. The church bells would ring at the local noon, when the sun was in the South.
Not a problem when the journey took several days, and you probably didn't have a watch anyway.
One of the relics of this age is a clock on the railways station in Bristol (IIRC), which has two minute hands, set about 12 minutes apart. One shows railway time (i.e. London time), and the other shows local time.
After a while it became obvious that simple timetables were more improtant than the sun being in the south at noon, and most places picked standard timezones.
I don't know when the US or France switched over though. Does anyone know how this came about?
IIRC The K7 does double-precision fp ops in a single cycle while the P-III requires 2 cycles. In which case, before interpreting these results we need to know how many of the SPEC tests use dp. The only test which mentions dp is 146.wave5. So I guess the margin for the K7 will only be slight lower for pure sp, but significantly higher for pure dp.
Also the results on the website are all 'BASE' values (i.e. little compiler optimisation). Presumably the P-III will beat the K7 under optimisation until compilers are modified for K7 optimisation.
Still, these figures are uncouraging.
The Gnome/Netscape problem and Debian
on
qt 2.0 released
·
· Score: 1
Very good point!
Linking Netscape with GTK is no problem, because the NPL allows it. However, linking any GPL'ed gnome apps with the NGLayout library would be exactly the same situation as linking a GPL'ed app (say KGhostView) with QT.
Of course the authors of such a Gnome app can resolve the problem by simply relicensing any app which calls NGLayout under a looser license.
The difficulty for the KDE developers is that some of their apps include GPL'ed code, and therefore cannot be relicensed without replacing the GPL'd code.
(The same problem *may* exist for some Gnome apps, but remember all KDE apps need QT, only a few Gnome apps use NGLayout).
It would be useful to have a list of all the KDE apps which include GPL code. It is possible that Debin could distribute KDE by simply throwing out some of the apps.
Kevin (running KDE 1.1.1 for now, but not KGhostView)
"Then you read about some of the functions and you smirk because you see the half-hearted attempts to emulate Unix or X Windows, you can see the
entomology and everything."
IIRC, it is 60,000 times faster than a P-II 350 if all you want to do is 60,000 4-bit additions. It might be reasonable at DES cracking too. But for running Quake, you're still better off with a P-II.
After all what we are really talking about is hanging all our icons in a corridor, and walking backwards or forwards along that corridor. We just add one dimension of extra space to the basic desktop. Probably more convenient than having a line of virtual desktops, but not much.
'Event Horizon' suggests a non-euclidean space, which is far more interesting. The problem remains how to ensure thing which you need together are nearby in space. The hyperbolic tree still seems better, as others have pointed out.
Re:What hardware/software does it use?
on
Digital VCRs
·
· Score: 1
The article implies that the hard drive is more of a problem than the video IO. I think Alan Cox has a good deal of video stuff under control.
I'm suprised the hard disk is an issue though - I would have thought that enough ram buffers would avoid any difficulties with reading and writing simultaneously.
Slashdot Mirror
One reason that there is a stigma attached to non-peer review and net based papers is that you can write any old garbage. I should know, my most cited paper (by a factor of 5) is in an electronic journal (It's not garbage, but it is informational rather than scientific peice).
Publication lists are often used by funding boards to asses the credibility of a proposal from an author. While less than ideal, this is necessary because it would be impractical for every funding board to review every aspect of every proposers work - they have neither the time, or the breadth of expertise. If they included non-peer-review publications, then there would be some people who like writing churning out heaps of garbage papers.
(I've reviewed some which are clearly rubbish. I've also asked for the editorial board to pick another reviewer when I disagree with the author concerned).
At the same time it would be wrong to deny that there are a few fields in which a viewpoint has acheived 'monopoly status', and alternate views cannot be expressed, although it is uncommon for all the journals in a field to be so dominated unless the field is very narrow. Experience shows that the situation usually resolves itself as the current generation of reviewers retire, although that is frustrating for individual authors at the time.
Given the balance between not abusing the nations tax $ on project with no scientific merit, and not holding up the progress of science, I think the balence in the current system is about right, even if it goes wrong in a few specific cases. If peer-reviewed electronic publishing can reduce the number of bad cases, I'm all for it.
I agree that peer-review is vital to filter out crackpots and commercial propaganda. But rather than the traditional editorial board approach, why not slashdot style moderation?
Give each reader and each paper a credibitlty score. The credibility of the reader is based on the average credilibity of each paper they have submitted. Each reader can give a credibility score to each paper they read. The score for the paper will be the submitted scores, weighted by the credibility of each reviewer.
Now it is easy to make foulups in a paper the first time round, so an initial submission could be made to an editorial area. People could add comments slashdot style, and the authors could use these to revise the paper for final submission.
When I started reading this, I hated it. But now I love it! If someone launches a journal in my field this way, I'll gladly submit a paper. Of course, there is no real need to have individual journals, if the database can be searched flexibly.
I agree with you in prinicle, but not in practice.
The GUI is important because these days anything with a GUI is comfortingly familiar, and anything
without is dangerous and frightening. But, the problem of not adding bloat before the bulk of the install is an important one.
I'm not sure perl is an issue, since I think a perl runtime is quite small and available for many platforms. But X, gtk and so on are certainly an issue.
To me the ideal approach seem to be a fake GUI which dresses up the CLI install - just like Stormix are doing. You can use a 16 colour VGA driver, which is small and will work on almost anything, and still draw good 'fake' gtk or qt widgets and window decorations. If it is done properly (and it looks like Stormix is), then the CLI option can be run with the same code, and I guess the bloat for the CLI can be kept to a fraction of a floppy.
I'm not sure whether this is the approach adopted by either Caldera or Stormix, or whether they are starting up a basic x-server. But its the way I'd do it.
I thought at first that writing it in perl was a bad idea, because you need a perl runtime environement (or does perl compile native these days?).
But one big advantage of using perl is that the port to Win32 should be fairly easy. Assuming the disk access code for windows can be ripped from FIPS and put in a library, the trickiest part will be the GUI. Of course, if they used a cross platform toolkit (either wxWindows or QT) then that would become a non-issue.
It is on X, but you can't tell X by looking. X is a protocol for talking to graphical terminals, just as VT100 was a protocol for talking to text terminals. This gets confusing with workstations, becuase the terminal and the host are the same computer.
Systems running X can look completely different, even Linux systems, depending on whether they use 75dpi or 100 dpi fonts, which window manager is running, which widget set and application uses, which theme the window manager and widget set is using (where applicable).
Things are further confused by look-alike themes. In some cases a Mac, or W95 theme running on X can be indistiguishable from the real thing.
There's very little point in it unless you have an interesting problem, and a heap of data.
But if you do, then there is suddenly a great deal of point. Unfortunately I don't think the machine they are proposing is very useful - it certainly wouldn't run any of my jobs, which will require a lot of work to migrate from vector supercomputers to any sort of parallel machine.
Even for jobs which can be massively parallelised, there is always the problem of compiling the results. Even for very simple tasks like searching for primes or crypto keys this becomes a headache with so many machines. For anything more complex, it will require a supercomputer with multiple fibre channels just to combine the results.
Where they are going to get a couple of megawatts to run this thing. Also, enough airconditioning to get rid of a couple of megawatts of heat.
Clearly the networking of this many machines will be impractical for most mathematical problems, I guess this will only be good for problems which could already be tackled by d.net or seti.
I think you've missed the point of the 200MHz system bus. The main benefit is not a faster path to the main memory, but sustained fast access to multiple paths, thus A DMA can be running at the same time as CPU access without a big slowdown.
Further, initial tests show the RAMBUS memory which will be able to keep up with faster system bus looks as though it will actually run slower in
most cases, because although the data rate is higher, the latency (the time taken to start a transfer) is much longer. Only very special applications will benefit, most will run slower.
Search www.theregister.co.uk for lengthy discussions of RAMBUS and its problems. PC133 looks like a better bet at the moment, but Intel has bet the house on RAMBUS.
Since a geosynch orbit has a radius of 42000 Km (22000 mi), and the Earth has a radius of 6400 Km (4000 mi), assuming zero axial tilt, a geosynch collector will be in sunlight for 22 hours 50 minutes each day.
Once you take into account the axial tilt, this will increase to very nearly 24 hours during the summer and winter or the ground station.
IIRC the reason microwave ovens work is that the microwaves are tuned to the energy of a bond in the water molecule. Thus microwaves heat up water. They don't heat up ice, and dry foods heat only slowly.
The problem is to get lots of energy from orbit to earth. Using mirrors and light would work, but light is absorbed by lots of things, and anything which absorbs lots of energy in any form is going to suffer.
So the challenge is to use a radiation which will go straight through almost anything that might stray into its path, but can still be picked up and converted by the receiving station. I guess the researchers have picked a band in the microwave spectrum which doesn't interact strongly with water or protein bonds - I don't know how easy this is. The receiver just needs an antenna designed for the particular band (this might be an array of simple rod-type dipoles, or a big horn/waveguide).
Of course (metal) airplanes will be a problem, but there are already plenty of no-fly-zones on the globe, I don't see a few more being a problem. (I assume the collector will be geostationary).
Well, you can use multiple network cards, but it makes the network run slower, not faster, hence the Mindcraft debacle. So I guess multiple NIC's aren't critical to a Beowulf.
Of course, the new threaded TCP/IP stack is now in the development version, so Linux should soon excell at benchmarks, and it might help cluster apps too.
The open source model is a close relative of the scientific development model, with code treated as research, and published for peer review and for the benefit of mankind.
And the problems are very similar. I've seen equivalent behaviour in the scientific community. Perhaps J was a little younger than many scientists - normally such disputes are played out in a manner more mature, although no less vicious.
Don't worry about it. Its human nature. The best we can do is remember to laugh at ourselves.
Jill Paton Walsh, Knowledge of Angels
It reminded me of Name of the Rose, without being heavy.But what I can do is compare it with other books. Now I admit that if you compare LotR with a lot of other fantasy, it comes off quite well.
It is more difficult to compare across genres, but I think you can compare universal elements such as prose style and characterisation. And when I compare the style with some mainstream novels, it seems somehow lacking:
- For an example of deep and moving characterisation, try 'Portrait of a Lady' by Henry James.
- For an example of beautiful prose, try '100 years of Solitude' by Gabriel Garcia Marquez.
- Iain Banks 'The Bridge' and 'Walking on Glass' paint vivid worlds. The latter has a gripping plot or three.
I don't think the literary style in LotR can seriously be compared with any of these.I wouldn't recommend the first two as fun books to read, the first is depressing and the second is very slow, but at the same time, both are unforgetable. The two Banks novels I mention because the first is science fiction(ish), and the second has fantasy elements.
What is the point I am trying to make? I'm not really sure. But it is something like this:
There are books you read and remember for the rest of your life. And there are books you grab to read on a plane because it is fun and passes the time, but once you throw them away they're forgotten.
Going back to LotR after 15 years and a lot of reading, it feels as though it should be in the second group. But it isn't. The only reason I can give is that it was my first fantasy, just as 'Islands in the Sky' was my first SF, and both have a special place, not for any particular technical merit, but because they were first.
But then again, I don't know anything about writing. All I can say is that I've read a bit of everything.
b) No. Apart from being first, its not very good. For a detailed critique from someone with more literary background than myself, look here.
It's a funny one this. I almost never swear, I'm put off by swearing (in real life), I'm a Christian. I'm frequently bored by stand up comics who think they can get a laugh by adding a few obscenities to an otherwise dull routine, and depressed to find that often they can.
South Park was the funniest movie I have ever seen.
I guess the issue might be that South Park pokes fun at the Moral Majority. If you're not part of it, and don't take yourself too seriously, then its funny.
Here's another odd thing: Sabrina, Buffy and Daria all feature attractive heroines who feel rejected by the fashionable set. Are they pandering to geek fantasies, or the inadaquacy/rejection thing is far more widespread, with personal significance to the majority of the population? I suspect the latter.
> I mean, it is simply waste of developers time.
If they were being paid, maybe so. But since most of the developers work for the satisfaction of it, I don't think we really have much say in the matter.
And of course, the competition leads to improvement and innovation.
Before the rise of the railways, it was normal for cities in Britain, and most probably everywhere else as well, to use their own local time. Thus, in traveling a few hundred miles east or west, you would have to reset your watch. The church bells would ring at the local noon, when the sun was in the South.
Not a problem when the journey took several days, and you probably didn't have a watch anyway.
One of the relics of this age is a clock on the railways station in Bristol (IIRC), which has two minute hands, set about 12 minutes apart. One shows railway time (i.e. London time), and the other shows local time.
After a while it became obvious that simple timetables were more improtant than the sun being in the south at noon, and most places picked standard timezones.
I don't know when the US or France switched over though. Does anyone know how this came about?
From www.3ag.net
CPU's at
500MHz $298
550MHz $459
600MHz $686
I guess that with several mobo vendors, long term mobo prices shouldn't be more than a current dual P-II, but the first few are bound to be pricey.
Also the results on the website are all 'BASE' values (i.e. little compiler optimisation). Presumably the P-III will beat the K7 under optimisation until compilers are modified for K7 optimisation.
Still, these figures are uncouraging.
Very good point!
Linking Netscape with GTK is no problem, because the NPL allows it. However, linking any GPL'ed gnome apps with the NGLayout library would be exactly the same situation as linking a GPL'ed app (say KGhostView) with QT.
Of course the authors of such a Gnome app can resolve the problem by simply relicensing any app which calls NGLayout under a looser license.
The difficulty for the KDE developers is that some of their apps include GPL'ed code, and therefore cannot be relicensed without replacing the GPL'd code.
(The same problem *may* exist for some Gnome apps, but remember all KDE apps need QT, only a few Gnome apps use NGLayout).
It would be useful to have a list of all the KDE apps which include GPL code. It is possible that Debin could distribute KDE by simply throwing out some of the apps.
Kevin (running KDE 1.1.1 for now, but not KGhostView)
IIRC, it is 60,000 times faster than a P-II 350
if all you want to do is 60,000 4-bit additions.
It might be reasonable at DES cracking too. But for running Quake, you're still better off with a P-II.
After all what we are really talking about is hanging all our icons in a corridor, and walking backwards or forwards along that corridor. We just add one dimension of extra space to the basic desktop. Probably more convenient than having a line of virtual desktops, but not much.
'Event Horizon' suggests a non-euclidean space, which is far more interesting. The problem remains how to ensure thing which you need together are nearby in space. The hyperbolic tree still seems better, as others have pointed out.
The article implies that the hard drive is more of a problem than the video IO. I think Alan Cox has a good deal of video stuff under control.
I'm suprised the hard disk is an issue though - I would have thought that enough ram buffers would avoid any difficulties with reading and writing simultaneously.