I do feel that attention to detail in one is reflected in the other and that overall quality will improve in neither until people start to care and it becomes less socially acceptable to make the mistake in the first place than to be the one to point the mistake out, in code or otherwise.
In my experience, you've got it backwards. Before I became a photographer I did embedded software for 20 years, shipping over 100M units and often having the final signature to begin fabricating my code into masked ROM. What I found was that overemphasis on "blame" instead of "results" was counterproductive. I seem to recall a discussion by Knuth on the point, but lack a citation.
Where you and I agree is on the idea that caring about the quality of ones code matters. It matters enormously, I've had the opportunity to primarily work with engineers who really do want to ship good, quality product. In the environments I've worked in, the occasional snarking at a bug has been counterproductive. It makes programmers defensive about their code, rather than being open to review and criticism, and thereby reduces the quality of the final product. Your experience may vary.
My point stands, though. In very large projects, it's easy for mistakes to slip through. Arbitrarily large projects will suffer some fraction of errors. The solution to this is not assuming that all software developers are lazy and ignorant (although some certainly are), instead, the solution is in better proecesses, testing, and tools.
*nods* It's not particularly difficult to get "almost no quality loss" from a standard image compression format, if all you are interested in is compressing an image and you're willing to tolerate "almost no quality loss."
However, RAW image files are, in many respects, a lot more than a pure image. They comprise the total of the image sensor readings and any white balance readings and/or seetings applied.
Why does this matter? A few examples should suffice.
Post-exposure white balance adjustment. In many applications a good setting for white-balance can not be convientently determined at processing time. In January 2004 I took a number of images at Mono Lake after dusk underneath a Pogonip layer, the appropriate "white light temperature" for that situation was extraordinarily blue. A conversion to a pure image format followed by an attempt to bring the colors back to natural would have failed in many cases because more signficant bits of the weak red and green channels remain. (Do note that the RGB sensors are not going to map 1:1 into image R,G,B channels, but you get the idea.) Score 1 point, RAW.
Noise reduction: In general, noise is going to be a function of which color sensor is being read. After applying filtering to the sensor results in order to undo the Bayer pattern sampling, you've lost a lot of ability to perform noise reduction, because you've lost the original sense of the staggered locations of the pixels. Another point, RAW.
Chromatic aberration: Same issue applies, slightly better CA corrections should be possible before the Bayer pattern interpolation. Game, set and match, RAW.
I've omitted a few more places where RAW image processing offers the quality-minded photographer better results than a conversion to an standard image file format. And that's before we get to the usual hyperbole about what "almost no quality loss" means (at least it's hyperbolic when you're trying to be able to produce top-quality enlargements.)
I've seen reference to a few companies who have reverse-engineered this new format, and of course Adobe chooses not to do that, are you aware of any product currently shipping willing to use Nikon's SDK?
Actually, the 1Ds can be had for about $4K now, if you know someone who will give me $5K for one I'd be happy to take it, and apply it to a purchase
of a 1Ds Mark II ($8K). (This nitpick doesn't change your point.)
I doubt you'll see a lot of switching from Nikon to Canon over this specific issue and Photoshop. Many professional photographers use other tools for the RAW conversion part of their workflow even if they use Photoshop later, for example, I use Phase One Capture One. I will be curious to find out how Phase One deals with the same issue.
Still, I often am faced with editing thousands of images at a time. Selecting the best images from a shoot is surprisingly difficult work, having the images come up by default with an incorrect white balance would, in fact, be a signficant hurdle.
Second, it's only white balance information. It's what the photographer told the camera about "white" or "gray" at the time of the shot, but it doesn't change the underlying image data. It's nothing that can't be recovered in the digital darkroom during processing.
If you shoot rarely, this might not seem an issue. If instead, like me, you shoot gigabytes of images a day for a week or two at a time and then have to sort through the results, the issue is larger. I agree it's not the end of the world, but it's not trivial either, and the tension between Adobe and Nikon on this point is interesting to try and make sense of, as well.
Doesn't appear to be an MIT puppets site, check the IPs--caltech.edu and caltechvsmit.com are in the same/16, and I'm pretty sure that Caltech at least at one point rated a class B address space.
Caltech alum here (Darb/Ma/84), I'm wondering if the parent is thinking of the Cannon thing (the Fleming Cannon, as I recall, was stolen back and forth with Mudd a few times--maybe as late as the 1970s?)
No, scratch that - I need a Powerbook. The one issue I have with it that keeps me from buying one is that the display is a lot smaller (1280x854) than what I use now on my ancient Dell (1600x1200).
I got the 1920x1200 (1280?). Very good sharpness, side-to-side hue consistency. The Apple screen is bright and lovely too, but even the 17" doesn't have the screen res. The Dell (15" at 1920 pixels) has almost unusably small pixels, but it works very, very well for the main purpose the laptop was purchased forinspecting photographic images during trips--I don't want to get back from a trip to Montana and find out a critical image isn't quite sharp or has some other problem (and yes, the extra resolution does help make that a more efficient process.)
I don't really "need" one or the other brand, the software I primarily use these days (weird, I feel like such a user, ha!) runs quite well on both platforms, and color management is now well-supported, by software or platform, on either base. But oh, that 17" powerbook, do I waaaaaaant it? Yeah. Just not $1100 and a lower res screen "yeah."
Looked into switching a bit more than a year ago for my last laptop purchase. For a best-monitor, 2GB RAM, 60 or 80GB disk configuration with as-comparable-as-possible other details, the Powerbook was $1100, or over a third again as expensive as the comparable Dell, and the Dell had, in some ways, a nicer display, higher-resolution although physically smaller. If it had been a few hundred, and I hadn't had to lose resolution, I might have gone through with it, but for me, last year, the Apple Premium was too high.
I agree with you about DejaNews, am optmistic about Keyhole (which Google dropped the price of
after acquisition), think their desktop tools rock. Just sayin'.
This page shows the dividing line, the probability needed to bring one from Torino 4 to Torino 5 depends on the kinetic energy as well. The current energy estimate I see at the NASA 2004 MN4 site is 1.6e3 MT, which puts us about (I'll guess here) a fifth of the way from 1e3 to 1e4 (as 1.6^5 = 10.48576, love them powers of two), so it's about 6/15ths (2/5ths) of the way between 1e2 and 1e5 MT, the upper bound of the Torino 4/5 box. It's kinda hard to figure how things transition on the probability scale there, but 2/5ths would clearly be to the right of the 10% mark. So I'll spitball that the dividing line is kinda sorta 15% at current energy levels. Your guess may vary.
Spammers are already doing a DDOS on me. Despite multiple spam filters, I've had to shut down domains for email entirely just to survive the onslaught.
Tax-and-spend politicians of every stripe, in fact. Just checked who signed that legislation into law, and which party passed that bill in the House. (I'm a whacko mix of strange poltiics, myself.)
But I agree with you about the body armor, big time.
Sure. The snark was in reference to the ten billion dollar payout to tobacco farmers, in "compensation" for them giving up other subsidies we'd given them in the past. As a US thing, it's not (and I mentioned this in a ccomment to my own comment) a fair alap against what would be a non-US $2B expenditure.
We always get our results in a timely manner and, to my recollection, there have never been any problems with the vote counting.
While I'm going to generalize here a bit, the problems that have come up with various paper voting systems usually reflect that ambiguously marked ballots, unless one is careful to define uniform acceptance criteria for marks on a ballot, can lead to "selective enforcement" and bias. This was the "chads in Florida" problem in 2000, but it can apply to most human-made marks.
Paper ballots also have a higher "pilot error" error rate, as it's (for example) possible to mark votes for more than one candidate accidentally.
Slashdot Mirror
I know, I know. :)
Yep. :)
In my experience, you've got it backwards. Before I became a photographer I did embedded software for 20 years, shipping over 100M units and often having the final signature to begin fabricating my code into masked ROM. What I found was that overemphasis on "blame" instead of "results" was counterproductive. I seem to recall a discussion by Knuth on the point, but lack a citation.
Where you and I agree is on the idea that caring about the quality of ones code matters. It matters enormously, I've had the opportunity to primarily work with engineers who really do want to ship good, quality product. In the environments I've worked in, the occasional snarking at a bug has been counterproductive. It makes programmers defensive about their code, rather than being open to review and criticism, and thereby reduces the quality of the final product. Your experience may vary.
My point stands, though. In very large projects, it's easy for mistakes to slip through. Arbitrarily large projects will suffer some fraction of errors. The solution to this is not assuming that all software developers are lazy and ignorant (although some certainly are), instead, the solution is in better proecesses, testing, and tools.
Using an interjection when you mean a adjectival phrase is an amateur mistake. Check your God-damned grammar.
Or they could have just Googled up the digits themselves, as they exist as files on the net, in locations such as this one.
*nods* It's not particularly difficult to get "almost no quality loss" from a standard image compression format, if all you are interested in is compressing an image and you're willing to tolerate "almost no quality loss."
However, RAW image files are, in many respects, a lot more than a pure image. They comprise the total of the image sensor readings and any white balance readings and/or seetings applied.
Why does this matter? A few examples should suffice.
Post-exposure white balance adjustment. In many applications a good setting for white-balance can not be convientently determined at processing time. In January 2004 I took a number of images at Mono Lake after dusk underneath a Pogonip layer, the appropriate "white light temperature" for that situation was extraordinarily blue. A conversion to a pure image format followed by an attempt to bring the colors back to natural would have failed in many cases because more signficant bits of the weak red and green channels remain. (Do note that the RGB sensors are not going to map 1:1 into image R,G,B channels, but you get the idea.) Score 1 point, RAW.
Noise reduction: In general, noise is going to be a function of which color sensor is being read. After applying filtering to the sensor results in order to undo the Bayer pattern sampling, you've lost a lot of ability to perform noise reduction, because you've lost the original sense of the staggered locations of the pixels. Another point, RAW.
Chromatic aberration: Same issue applies, slightly better CA corrections should be possible before the Bayer pattern interpolation. Game, set and match, RAW.
I've omitted a few more places where RAW image processing offers the quality-minded photographer better results than a conversion to an standard image file format. And that's before we get to the usual hyperbole about what "almost no quality loss" means (at least it's hyperbolic when you're trying to be able to produce top-quality enlargements.)
I've seen reference to a few companies who have reverse-engineered this new format, and of course Adobe chooses not to do that, are you aware of any product currently shipping willing to use Nikon's SDK?
I doubt you'll see a lot of switching from Nikon to Canon over this specific issue and Photoshop. Many professional photographers use other tools for the RAW conversion part of their workflow even if they use Photoshop later, for example, I use Phase One Capture One. I will be curious to find out how Phase One deals with the same issue.
Still, I often am faced with editing thousands of images at a time. Selecting the best images from a shoot is surprisingly difficult work, having the images come up by default with an incorrect white balance would, in fact, be a signficant hurdle.
If you shoot rarely, this might not seem an issue. If instead, like me, you shoot gigabytes of images a day for a week or two at a time and then have to sort through the results, the issue is larger. I agree it's not the end of the world, but it's not trivial either, and the tension between Adobe and Nikon on this point is interesting to try and make sense of, as well.
Nevermind, I had misread the parent of my post (parent). Mod down both at will, I should have read more carefully.
Doesn't appear to be an MIT puppets site, check the IPs--caltech.edu and caltechvsmit.com are in the same /16, and I'm pretty sure that Caltech at least at one point rated a class B address space.
Caltech alum here (Darb/Ma/84), I'm wondering if the parent is thinking of the Cannon thing (the Fleming Cannon, as I recall, was stolen back and forth with Mudd a few times--maybe as late as the 1970s?)
Absolutely. (Caltech alum, myself.)
I got the 1920x1200 (1280?). Very good sharpness, side-to-side hue consistency. The Apple screen is bright and lovely too, but even the 17" doesn't have the screen res. The Dell (15" at 1920 pixels) has almost unusably small pixels, but it works very, very well for the main purpose the laptop was purchased forinspecting photographic images during trips--I don't want to get back from a trip to Montana and find out a critical image isn't quite sharp or has some other problem (and yes, the extra resolution does help make that a more efficient process.)
I don't really "need" one or the other brand, the software I primarily use these days (weird, I feel like such a user, ha!) runs quite well on both platforms, and color management is now well-supported, by software or platform, on either base. But oh, that 17" powerbook, do I waaaaaaant it? Yeah. Just not $1100 and a lower res screen "yeah."
Looked into switching a bit more than a year ago for my last laptop purchase. For a best-monitor, 2GB RAM, 60 or 80GB disk configuration with as-comparable-as-possible other details, the Powerbook was $1100, or over a third again as expensive as the comparable Dell, and the Dell had, in some ways, a nicer display, higher-resolution although physically smaller. If it had been a few hundred, and I hadn't had to lose resolution, I might have gone through with it, but for me, last year, the Apple Premium was too high.
Perhaps you misunderstand Google's "main goal."
I agree with you about DejaNews, am optmistic about Keyhole (which Google dropped the price of after acquisition), think their desktop tools rock. Just sayin'.
This page shows the dividing line, the probability needed to bring one from Torino 4 to Torino 5 depends on the kinetic energy as well. The current energy estimate I see at the NASA 2004 MN4 site is 1.6e3 MT, which puts us about (I'll guess here) a fifth of the way from 1e3 to 1e4 (as 1.6^5 = 10.48576, love them powers of two), so it's about 6/15ths (2/5ths) of the way between 1e2 and 1e5 MT, the upper bound of the Torino 4/5 box. It's kinda hard to figure how things transition on the probability scale there, but 2/5ths would clearly be to the right of the 10% mark. So I'll spitball that the dividing line is kinda sorta 15% at current energy levels. Your guess may vary.
Spammers are already doing a DDOS on me. Despite multiple spam filters, I've had to shut down domains for email entirely just to survive the onslaught.
But I agree with you about the body armor, big time.
Sure. The snark was in reference to the ten billion dollar payout to tobacco farmers, in "compensation" for them giving up other subsidies we'd given them in the past. As a US thing, it's not (and I mentioned this in a ccomment to my own comment) a fair alap against what would be a non-US $2B expenditure.
(And yes, yes, different countries, unfair comparison.)
Yeah, because we have to spend five times that much to give payola to tobacco farmers. Not that I'm bitter.
Excellent point.
While I'm going to generalize here a bit, the problems that have come up with various paper voting systems usually reflect that ambiguously marked ballots, unless one is careful to define uniform acceptance criteria for marks on a ballot, can lead to "selective enforcement" and bias. This was the "chads in Florida" problem in 2000, but it can apply to most human-made marks.
Paper ballots also have a higher "pilot error" error rate, as it's (for example) possible to mark votes for more than one candidate accidentally.