Do 'Western Digital' and 'server' belong in the same sentence? I mean people would be a bit bemused if you could buy eMachines 'enterprise-class' hardware or subscribe to 'AOL Datacenter Edition'.
Is the popular view of WD drives wrong? Or are all the manufacturers just as bad these days in the consumer space?
I don't mind having configuration options, but IMHO they should follow two fundamental rules:
- The default configuration should be usable for most users (and particularly for new users). Havoc Pennington's article talks about the 'unbreak my application please' configuration option for Emacs's clipboard handling. There should not be any need to change preferences just to make apps work correctly and especially not just to make behaviour consistent between applications.
- Configure something only *once*. If you're changing something like double-click sensitivity or dialogue box fonts or keyboard accelerators, it shouldn't be necessary to set these things separately for different apps. KDE and GNOME go a long way towards this but still it would be better if the control panel of KDE could set configuration for GNOME applications, at least for basic stuff. Any case where the user has to configure the same preference twice should be considered a bug.
Yes, you can harvest addresses from your referrer log if people come from Hotmail to your site. At least, you could: I haven't checked if Hotmail's URLs still include the email address or some munged version of it.
Yeah, it's a shame that XFree86's monitor database doesn't include the physical dimensions of the monitor so it can set the dpi correctly. (Although I would prefer to use micrometres-per-dot rather than dots-per-inch.)
Are there no HDTV tuner cards for PCs? If they were available you could watch HDTV on your existing monitor and probably it would be a lot cheaper than $6k.
OK. I was really just talking about experience with a Philips CM8833 monitor, which is really a TV set without tuner. On that I could get 768x288 (more, in fact, if you're prepared to make the pixels even narrower than 1:2).
A pixel is never a point anyway, it is either a square (on LCD monitors, or CRTs with a high enough dot pitch relative to the display resolution) or a round fuzzy thing (on low-end CRTs).
The movie screens would not have become long and thin if moviegoers didn't prefer it. Unless we have all been hoodwinked by the movie industry into 'preferring' an inferior format. Why, for example, didn't they make the screens tall and narrow to generate some marketing advantage over television? There must be _some_ reason to have a screen a lot wider than it is tall.
FWIW, isn't a movie screen a lot wider than the 'widescreen' televisions? A widescreen TV is 16:9 but the last time I went to the cinema it seemed rather more than that. (Who knows, perhaps they are using the golden ratio, which Open University presenters tell us is the most 'natural' aspect ratio for an image, without giving any evidence.)
I agree that widescreen monitors are usually quite stupid, especially if your work consists mostly of editing text in one form or another. But manufacturers need to have their spurious 'multimedia' sales pitches, even if they don't use that word any more.
It should be possible to turn your monitor on its side, if it is an LCD, and hack XFree86 to get a 960x1600 desktop or whatever. I remember seeing old Macs with displays taller than they were wide, used for viewing a whole page in DTP I think. I wonder if the new Mac displays can be rotated to get the same effect. Somehow, DTP isn't as sexy as it once was.
Is that 640x480 interlaced at 60Hz, or non-interlaced? Does it include the overscan area?
I'm curious because I'm not familiar with TV sets in other countries. In Britain the PAL, 50Hz system can get 768x288 non-interlaced, if you use some of the overscan area. I think you could interlace the picture and go to 768x576, but it looks _awful_.
Most TV sets have some kind of broken analogue 'sharpening' filter to try and de-smear the image. I wonder if you could design a set of fonts that takes advantage of this, so that the image, once it has been downgraded to the low bandwidth of the television and then 'sharpened' again, is a reasonable approximation to the original. Or you could even try to do some DSPing on your images to get the optimum television display quality based on some assumption of what the TV will do to sharpen the image.
(Obviously not all images could be displayed perfectly on the TV, there would be some loss of detail. Ye cannae change the laws of physics nor those of information theory. But there might be some class of images which can be sent down the wire to the television, sharpened at the other end and come out looking okay.)
What's wrong with non-square pixels? Why should it matter if you are sending television pictures? For that matter, why should square pixels be so important for computer displays either? (I know that they _are_ right now, because we still have loads of windowing systems using pixmaps which have been created for square pixels, but with SVG icon support and scalable fonts there's no reason why the display couldn't adapt to whatever pixel size you have.)
Not that I'm advocating changing the square pixel to something else for computer displays - but I don't see why it's such a stupid idea just for televisions.
But anyway, do television sets even have pixels? I thought they had scan lines and within each scan line there were no discrete pixels, just a continuous smear of different colours.
Teletext is only 40x25 characters IIRC, even a DOS text-mode program expects 80 columns. Make the characters any narrower and they will not be particularly easy to read, even with specially designed glyphs.
In PC and server environments, Phoenix cME enables the creation and management of a secure "host protected area" (HPA) of the hard drive, where applications reside. The first of these are Phoenix's own FirstWave apps, which help diagnose and recover PCs if the OS goes tits-up, third party developers can write their own aps for storage in this "tamper-proof"(actually tamper-resistant) area.
This sounds very much like the 'reference partition' used with second-generation (1991-1995) PS/2s. The older models had a 'reference diskette'... but it's too much to ask nowadays that manufacturers could write a diagnostic program to fit in a mere 1440 kilobytes.
You should try Dillo, it is screamingly fast even on slow machines. That, xterm, icewm and xemacs should be all the X11 applications you ever need to run anyway;-).
You're right, you can make an explicit copy to emulate value semantics, but it's inconsistent with what happens for other data types. For example
$a = 55; $b = $a; $b = 66; print $a;
Clearly the original value of $a has not been changed. You don't need any voodoo with square brackets or reading 'man perltoot' to get this. And if you do the same thing with lists, it's fine too:
And bam, it's different. You don't have to take references to strings to store them in hashes, why should you have to take references to lists to store them? If I want a hash of lists, why can't Perl manage it?
Of course you should be able to make explicit references if you want, but they shouldn't be forced on you just for simple things like the above.
'Easy things should be easy' - I wish Perl would follow its own motto in this area. It's good that there is documentation such as perltoot to guide the new programmer through this stuff, but rather odd that none of it seems to acknowledge the possibility that this stuff is anything other than bleeding obvious.
The problem (Windows shows the splash screen, but then BSODs) sounds like it could have been caused by using the Linux NTFS driver in r/w mode. I broke a machine once by doing that, I thought that just copying a single file across would be okay, but it wasn't. I knew the write support in ntfs was experimental but hadn't realized it was quite _that_ experimental...
No, Cyrix was sold to VIA, but VIA isn't using the Cyrix design any more. They released a chip based on the Cyrix stuff originally, but it sucked. Then they bought Centaur as well, and the current VIA C3 is based on Centaur's WinChip family and made by the same design team.
Whoever made the Pentium 133 comment is being silly. Such a machine is more than adequate to run a current Linux distribution and a good range of applications (Emacs, TeX, web browsing with Netscape 4 or Dillo, KDE or GNOME if it has enough RAM, even development work in C or scripting languages).
If Microsoft wanted to be relevant in the future, she told them, it had to adjust to NetGen, even if it meant producing software that the middle-aged guys in the room didn't care for.
In early 2001, she set up in the hip waterfront area of Seattle--miles away from the orthodoxy at Redmond--recruiting kids barely out of college, promising them the opportunity to make an immediate impact.
Is it me, or does the whole article feel five years out of date?
Threedegrees is a surprising departure for Microsoft. The company that's relentlessly focused on productivity has now produced an anti-productivity tool, constantly interrupting you and urging you to waste time with your friends.
Has this writer never heard of the Office Assistant? Or the stupid 'DHTML' scripting features which IE is so keen to support? Or window animations, or sound effects, or screensavers... the whole focus of Microsoft software is to get something that looks cool before something that works and gets out of your way.
What I was asking is how much of this rendering is (a) done in hardware on the video card and (b) so slow that a newer card could shave 30 seconds off the time.
30 seconds off a complex render? Surely nothing rendered *in hardware*, on the video card, is that slow. For software rendering I can believe it.
I thought the point of 3d acceleration in video cards is to give a real-time display, and for that obviously faster is better. If you are doing massive rendering jobs a la Toy Story, those are done in software. It would be interesting to accelerate that task using standard 3d accelerator cards, but I think the drivers and/or hardware tend to have bugs which mean missing surfaces, etc, cannot be ruled out. So they may not be good enough to generate the final version.
Slashdot Mirror
Do 'Western Digital' and 'server' belong in the same sentence? I mean people would be a bit bemused if you could buy eMachines 'enterprise-class' hardware or subscribe to 'AOL Datacenter Edition'.
Is the popular view of WD drives wrong? Or are all the manufacturers just as bad these days in the consumer space?
I don't mind having configuration options, but IMHO they should follow two fundamental rules:
- The default configuration should be usable for most users (and particularly for new users). Havoc Pennington's article talks about the 'unbreak my application please' configuration option for Emacs's clipboard handling. There should not be any need to change preferences just to make apps work correctly and especially not just to make behaviour consistent between applications.
- Configure something only *once*. If you're changing something like double-click sensitivity or dialogue box fonts or keyboard accelerators, it shouldn't be necessary to set these things separately for different apps. KDE and GNOME go a long way towards this but still it would be better if the control panel of KDE could set configuration for GNOME applications, at least for basic stuff. Any case where the user has to configure the same preference twice should be considered a bug.
It's not a joke that's worth 'saving'...
Disappoiningly, not one of his answers begins 'IAAL, but...'
Yes, you can harvest addresses from your referrer log if people come from Hotmail to your site. At least, you could: I haven't checked if Hotmail's URLs still include the email address or some munged version of it.
Yeah, it's a shame that XFree86's monitor database doesn't include the physical dimensions of the monitor so it can set the dpi correctly. (Although I would prefer to use micrometres-per-dot rather than dots-per-inch.)
Like I said, a pixel is either a square or a fuzzy round thing. It is not a point. Pixels have area.
Are there no HDTV tuner cards for PCs? If they were available you could watch HDTV on your existing monitor and probably it would be a lot cheaper than $6k.
OK. I was really just talking about experience with a Philips CM8833 monitor, which is really a TV set without tuner. On that I could get 768x288 (more, in fact, if you're prepared to make the pixels even narrower than 1:2).
A pixel is never a point anyway, it is either a square (on LCD monitors, or CRTs with a high enough dot pitch relative to the display resolution) or a round fuzzy thing (on low-end CRTs).
The movie screens would not have become long and thin if moviegoers didn't prefer it. Unless we have all been hoodwinked by the movie industry into 'preferring' an inferior format. Why, for example, didn't they make the screens tall and narrow to generate some marketing advantage over television? There must be _some_ reason to have a screen a lot wider than it is tall.
FWIW, isn't a movie screen a lot wider than the 'widescreen' televisions? A widescreen TV is 16:9 but the last time I went to the cinema it seemed rather more than that. (Who knows, perhaps they are using the golden ratio, which Open University presenters tell us is the most 'natural' aspect ratio for an image, without giving any evidence.)
I agree that widescreen monitors are usually quite stupid, especially if your work consists mostly of editing text in one form or another. But manufacturers need to have their spurious 'multimedia' sales pitches, even if they don't use that word any more.
It should be possible to turn your monitor on its side, if it is an LCD, and hack XFree86 to get a 960x1600 desktop or whatever. I remember seeing old Macs with displays taller than they were wide, used for viewing a whole page in DTP I think. I wonder if the new Mac displays can be rotated to get the same effect. Somehow, DTP isn't as sexy as it once was.
Is that 640x480 interlaced at 60Hz, or non-interlaced? Does it include the overscan area?
I'm curious because I'm not familiar with TV sets in other countries. In Britain the PAL, 50Hz system can get 768x288 non-interlaced, if you use some of the overscan area. I think you could interlace the picture and go to 768x576, but it looks _awful_.
Most TV sets have some kind of broken analogue 'sharpening' filter to try and de-smear the image. I wonder if you could design a set of fonts that takes advantage of this, so that the image, once it has been downgraded to the low bandwidth of the television and then 'sharpened' again, is a reasonable approximation to the original. Or you could even try to do some DSPing on your images to get the optimum television display quality based on some assumption of what the TV will do to sharpen the image.
(Obviously not all images could be displayed perfectly on the TV, there would be some loss of detail. Ye cannae change the laws of physics nor those of information theory. But there might be some class of images which can be sent down the wire to the television, sharpened at the other end and come out looking okay.)
What's wrong with non-square pixels? Why should it matter if you are sending television pictures? For that matter, why should square pixels be so important for computer displays either? (I know that they _are_ right now, because we still have loads of windowing systems using pixmaps which have been created for square pixels, but with SVG icon support and scalable fonts there's no reason why the display couldn't adapt to whatever pixel size you have.)
Not that I'm advocating changing the square pixel to something else for computer displays - but I don't see why it's such a stupid idea just for televisions.
But anyway, do television sets even have pixels? I thought they had scan lines and within each scan line there were no discrete pixels, just a continuous smear of different colours.
Teletext is only 40x25 characters IIRC, even a DOS text-mode program expects 80 columns. Make the characters any narrower and they will not be particularly easy to read, even with specially designed glyphs.
What size characters does WebTV manage to use?
You should try Dillo, it is screamingly fast even on slow machines. That, xterm, icewm and xemacs should be all the X11 applications you ever need to run anyway ;-).
You're right, you can make an explicit copy to emulate value semantics, but it's inconsistent with what happens for other data types. For example
$a = 55;
$b = $a;
$b = 66;
print $a;
Clearly the original value of $a has not been changed. You don't need any voodoo with square brackets or reading 'man perltoot' to get this.
And if you do the same thing with lists, it's fine too:
@a = (55, 56);
@b = @a;
push @b, 'hello';
print join(', ', @a);
Again the original object is unchanged. Now, what if instead of variables I have another data structure (say, a hash) storing these values?
$h{a} = (55, 56);
Well there's the first problem, you can't do that, it has to be a scalar. Well, okay,
$h{a} = [ 55, 56 ];
$h{b} = $h{a};
push @{$h{b}}, 'hello';
print join(', ', @{$h{a}});
And bam, it's different. You don't have to take references to strings to store them in hashes, why should you have to take references to lists to store them? If I want a hash of lists, why can't Perl manage it?
Of course you should be able to make explicit references if you want, but they shouldn't be forced on you just for simple things like the above.
'Easy things should be easy' - I wish Perl would follow its own motto in this area. It's good that there is documentation such as perltoot to guide the new programmer through this stuff, but rather odd that none of it seems to acknowledge the possibility that this stuff is anything other than bleeding obvious.
The problem (Windows shows the splash screen, but then BSODs) sounds like it could have been caused by using the Linux NTFS driver in r/w mode. I broke a machine once by doing that, I thought that just copying a single file across would be okay, but it wasn't. I knew the write support in ntfs was experimental but hadn't realized it was quite _that_ experimental...
No, Cyrix was sold to VIA, but VIA isn't using the Cyrix design any more. They released a chip based on the Cyrix stuff originally, but it sucked. Then they bought Centaur as well, and the current VIA C3 is based on Centaur's WinChip family and made by the same design team.
Whoever made the Pentium 133 comment is being silly. Such a machine is more than adequate to run a current Linux distribution and a good range of applications (Emacs, TeX, web browsing with Netscape 4 or Dillo, KDE or GNOME if it has enough RAM, even development work in C or scripting languages).
What I was asking is how much of this rendering is (a) done in hardware on the video card and (b) so slow that a newer card could shave 30 seconds off the time.
30 seconds off a complex render? Surely nothing rendered *in hardware*, on the video card, is that slow. For software rendering I can believe it.
I thought the point of 3d acceleration in video cards is to give a real-time display, and for that obviously faster is better. If you are doing massive rendering jobs a la Toy Story, those are done in software. It would be interesting to accelerate that task using standard 3d accelerator cards, but I think the drivers and/or hardware tend to have bugs which mean missing surfaces, etc, cannot be ruled out. So they may not be good enough to generate the final version.