This is part of the subtext both of the original article, and of this most recent post, so I thought I'd share what I know about it. FWIW, I'm a radiologist--that is, an MD who interprets the results of imaging studies--and an informatics geek.
Images are created on whatever imaging device--CT scanner, MR scanner, ultrasound machine, digital X-ray machine--and manipulated by the device's controlling system to do simple annotations, reformatting, etc. This is typically a Unix-based system running custom software designed and maintained by the device's vendor. The images are not usually interpreted on these systems.
There tend to be two front-ends for looking at images in the PACS database. The first is the radiologist's interface, which is a high-end video workstation dedicated to showing medical images with the greatest possible fidelity. Most systems I've seen are Windows-based (Windows 2000, in our case) and run software which was built by the the imaging system vendors in the late 1990's. Much is made of the "lossless" nature of the images which are displayed; for example, when you log into such a machine, you're warned about how "This is a medical device" and that you shouldn't mess with it. Much is also made of "diagnostic-quality monitors" and high-end video cards to drive the monitors. This is an artifact from the early days of digital imaging interpretation in radiology, when there was a great deal of concern about whether the quality of the digital images would be adequate for us to figure out what was going on in Grandma's chest X-ray if we weren't looking at a piece of acetate. Most of these concerns have died away, as the differences in resolution and dynamic range turned out to be relatively minor and the added conveniences of being able to manipulate the images digitally turned out to be huge. For example, the new LCDs I seen being put on PACS workstations are off-the-shelf Dell 22-inchers, as far as I can tell.
Finally, there are "non-diagnostic" interfaces to the PACS images, which do tend to be web-based. These are so non-radiologist doctors can look at the images, too. Some are IE-based, and use an ActiveX control to display the images, and some use a Java applet. These are displayed with lossy compression (since someone might want to look at them from off-site via a VPN), and officially are not allowed to be used for interpretation. And in fact, I wouldn't want to; it's a lot harder to see subtle things on them than on a full-blown PACS workstation. Part of that is just the interface (it's hard to use those stupid ActiveX/applet things) and part of it is crummy/mis-configured monitors, but I suppose compression artifacts could also play a role.
So, to review: you go see your doctor, Dr. Smith, in her office, and she orders a chest X-ray for you because you're coughing and have a fever. You come to the hospital, and the nice technologist takes frontal and lateral view of your chest on the digital X-ray machine. He then goes back to the X-ray control room, and sees that the images are pretty good, and so he sticks your name on them, and a marker of the date/time and his name, and so on, and then sends them to the hospital's PACS system. I (the radiologist) am working at my PACS workstation, going through the long list of all of the CT scans, MR scans, and X-rays taken in the hospital. I get to your chest X-ray and look at it; I don't seen any sign of pneumonia, so I write a report (the subject of a whole different set of informatics) that basically says "Clear lungs" and that gets entered into your electronic medical record. Then, Dr. Smith back in her office can see your X-ray via her Web-based interface. If she wonders about something she sees, she can call me up and say, "What's that stuff at the left ape
In my philosophy, the entire purpose of government is to protect individuals' rights. For example,
Protect its citizens from foreign threats
Prevent its citizens from using force on one another.
Prevent its citizens from defrauding one another.
Prevent its citizens from enslaving one another.
In my view, a "corporation" should have no rights of its own. (I'm not even that crazy about the idea of liability avoidance that's supposed to be the whole idea of incorporation, but that's another story.) So it goes without saying that the government should be in the business of preventing corporations from screwing individuals.
I'm not a huge extremist, but I do become disgusted with the extent to which the US government, far from protecting its citizens from mega-corporations, eagerly helps them exploit its citizens.
Re:Economic reasons to scare John Q. Public
on
What, Me Worry?
·
· Score: 1, Offtopic
I don't think the government would find that sufficiently satisfying. Such a "war" does not require massive invasions of civil liberties or decreases in government transparency in the name of "national security" to fight.
I'm betting we'll stick with our current "war" on terrorism.
In my experience, people who study science are much more broadly educated than people who don't (which seems obvious). But physics majors I've known have non-science interests (frequently music, interestingly) at a much higher rate than English majors have science-related interests. (Actually, have you ever noticed that many non-scientists consider technical stuff to be beneath them?)
As for myself, I studied biology, physics, and electrical engineering as an undergrad; I'm getting an MD and a PhD in neuroscience. As an undergrad, I took classes in history, political science, writing, and American literature. By coincidence, I was also heavily concentrated as an undergrad in theatre arts, so I do happen to know a little Shakespeare. And I still maintain an interest in almost all of those things, and regularly read about them.
But I don't consider myself unusualy well-rounded among others in science I encounter; on the contrary, I'm frequently impressed by the other interesting things that they do.
After reading the analysis, two things sprang out at me. The first is that a lot of the stuff on a Linux system is meant for development, rather than just using the system. The second is that lots of the stuff on the list clearly is "application" and not anyone's idea of an "operating system".
Specifically, in the top ten, we have:
Development Tools
gcc (#4)
gdb (#5)
binutils (#6)
Applications
emacs (#7)
LAPACK (#8)
gimp (#9)
mysql (#10)
(Also in the top 20 are libgcj, teTeX, postgresql, and xemacs. And we won't get into the issue of whether Mozilla (#2) should be considered part of the operating system.)
So my question is, what's the size of the non-development/non-application stuff? What's the size of the kernel plus the essential utilities (most of which are GNU, as RMS points out ad nauseum)?
I found something interesting in the preliminary injunction. The judge is drawing his conclusions:
1. The Court concludes that Plaintiffs are likely to succeed on their claims that:
... [a--e skipped] ...
(f) The Act violates the First and Fourteenth Amendments of the United States Constitution because it prevents people from communicating and accessing information anonymously.
Wow! There's a federal judge that seems to think the Constitution says that we should all be able to surf and e-mail anonymously! I'd like the Constitution to say that, but I don't think it does.
...is a wonderful interviewer. She really seems to understand enough about the relevant issues to ask intelligent questions, and she has a knack for making her guests comfortable enough to reveal themselves. The show's format helps, too; plenty of time for each interview (20 or 30 minutes), and no morons calling in.
Should be a really worthwhile listen; probably one of the best interviews Linus will do on his book tour.
AOL thinks there are lots of ways to make money on messaging. One of their executives is quoted as saying that Instant Messaging is AOL's most important asset.
But clearly, they think they can only succeed if they control the client as well as the servers. And a route into their system where they don't control the client is apparently unacceptable to them.
Their privacy policy is quite funny. It says, "Digital:Convergence wants every member of its community to understand what information we collect and how we intend to use that information."
But while the rest of the page is quite explicit about not giving out your phone number, it says nothing about the mother lode of data they (surely) collect about what you actually scan. In fact, they don't even mention the possibility!
...has nothing to do with IP or any other crap like that.
It seems pretty clear to me that this company is all about collecting personal information about you. Now, we're used to companies like Doubleclick et al. using various schemes to figure out where you browse, and generate models of your on-line behaviors automatically. But the CueCat thing is designed so that they can find out about things in the real world, too. They WANT you to scan things in your house (books, products, magazines, TV shows, etc), and when you do, their software does a quick lookup for you, and gives you the shovelware connected to whatever it is.
But it ALSO remembers who scanned what, and they can generate a much more impressive profile of you as a consumer: what do you watch, what magazines are you reading, what books do you have, etc. And I'll bet that THAT's what their business model is all about---more invasive consumer profiling.
And THAT's why they can't tolerate the Linux drivers which don't do the lookup through their servers---because they don't let them form that profile. They've probably calculated their (hugely expensive) giveaway scheme to give a particular rate of data generation through their servers, and if they don't get that, then their business model is shot to hell. Of COURSE they're going to try to prevent people from getting around their software: every person using the Linux driver is the same as a reader thrown away to them. So I'm sure that if a developer tried to work with them, they'd spend a lot of effort urging them to direct all connections through the Digital Convergance servers.
My father-in-law gave me his reader (which some magazine helpfully mailed him), and I was going to use it under Linux, but I've changed my mind. If I want a barcode reader, I'll go buy one that's free of this kind of crap.
It's another one of those prickly problems, I guess, if the law is going to have that little disregard for the idea of "responsibility". But in the end, it sounds like it's going to come down to some kind of monitoring scheme, right?
And maybe I'm hopelessly naive, but it seems like what you need to do is find a way to involve your employees in the process. Tell them what the problem is. Ask what kind of monitoring they'd find acceptable, and what they'd find intrusive. When you figure out what precisely your options are, ask your employees what they think of all of the different possibilities.
And when you figure something out, tell people as EXACTLY as you can what you're doing, and how it works. Tell them what you expect of them; tell them what limits you are going to respect (and then respect them, of course).
I'm sure some people will be a little put out, no matter what you come up with, but given that one can get all kinds of free e-mail accounts and that most have abundant access to the net, some restrictions on what one does at work shouldn't be too onerus.
What would it be illegal to stand on the border and shout into France?
Or, even more accurately, what would it be illegal to say over the telephone in France, regardless of where you are?
Or, yet more precisely, what would it be illegal for you to put as the outgoing message on your answering machine that anyone in France might call and hear?
But I don't think that's bad; in fact, it makes my inner libertarian quite happy. That's what makes the web so great! I can put anything I want on my site, and you can put anything you want on your site (including any URL you like).
And of course, it goes without saying that just because you ask me for a page I don't have to give it to you; I might put conditions on that. (Like charging you for it, or making you look at an ad or a silly splash screen.)
The only way there could be a legal recourse is if the government has the right to tell me how my site has to look and work, and I don't think anyone wants that.
Taxation on internet transactions is inevitable. We can argue over whether it's a good thing or not; an interesting contribution to the debate is Paul Krugman's oped piece in the NY Times (free registration, blah, blah, blah).
It seems to me, though, that the reason it isn't already taxed is not this garbage about strangling the new economy but the fact that it's the states that want to collect the taxes, and the Constitution specifically prohibits that. The politicians are just pretending to care about the Internet until they can figure out a way to tax it that's not against the constitution.
Seriously, though, at least in the Red Hat RPM they distributed, they seem to have hard-coded in the assumption that you're using Netscape. In fact, when you install the RPM, it tries to start Netscape and connect to the Real site (even though one usually is root for installing the RPM). Not cool...
And there isn't apparently any place to indicate an alternate browser to be used...
I downloaded the player this morning, and installed it---it's quite sweet. Works great with ESD, sounds very good. I'm quite pleased with it.
But there's an odd trend I've noticed... When I went to listen to the live stream of my NPR station with my spiffy new player, it couldn't play it because it was for Windows Media Player only. Doh! I wrote them to tell them that this was pretty frustrating (especially since I'm pretty sure they also used to webcast for Real player).
But the fact is, this is becoming more and more common, I think. Sites are going to WMP-only based content. Is Microsoft trying to undercut their competition again by bundling software with the OS? (And is the DoJ paying attention?)
I think this is the thrust of the argument against the Clinton administration's idea, because it would polarize the divide you describe even further.
At least now, even non-hackers can get pseudo-privacy (which, while not absolute, is better than nothing). This proposal would make a bandit out of anyone who tries to be anonymous. And does anyone think that those who would engage in questionable practices will have much problem getting around whatever enforcement is put in place? (How does that silly NRA slogan go? "When you outlaw anonymity, only the outlaws will be anonymous.")
In response to your proposal (which is interesting), how would the anonymity it creates be fundamentally different from the illusion we have now?
-- Happy Premise #3: Even though I feel like I might ignite, I probably won't.
It seems to me that "meme" brings with it an interesting connotations: (1) Memes spread; (2) Memes compete with one another; (3) People study the nature of (1) and (2).
So it seems like it might be a useful shorthand for "an idea spreading in competition with others".
It's a fairly bogus idea (in both senses, if you'll pardon the expression).
We're able to encode (and therefore to transmit and reconstruct) visual images digitally because we understand the relevant paramaters of digital images: intensity vs. two spatial dimensions and wavelength. Sound is even easier: intensity vs. time. It's even easier for these because to a reasonable approximation, these parametres are linear.
But we have NO idea of what the relevant parameters of a smell are. People have suggested all kinds of ideas over the years, but nothing has been really satisfactory. You've probably heard about the seven primary scents; that's pretty much a load of garbage, debunked by people who studied olfaction in the fifties and sixties. More recently, people have tried to understand the sense of smell in terms of the shapes of the molecules being detected, but even that has been unsatisfactory.
Perfumers, therefore, remain essentially what they were two hundred years ago: empiricists surrounded by a LOT of bottles. Hundreds of them. It's amazing to see. It takes YEARS to train someone to be good at it, and even then, their ability to reproduce an arbitrary odor is VERY limited. The notion that someone could put that in a box and sell it to people is absurd.
What you're much more likely to get is an odor equivalent of Jelly Belly jelly beans. You know how on the jelly beans, they say, "Combine cinnamon and popcorn to get the flavor of cinnammon popcorn balls"? Well, they'll choose a handful of cheap, fake scents, and use those over and over again. So you'll get things like "flowers" and "fruit" or even "fruity flowers", but never "the beach in Tahiti after a rainstorm".
Slashdot Mirror
This is part of the subtext both of the original article, and of this most recent post, so I thought I'd share what I know about it. FWIW, I'm a radiologist--that is, an MD who interprets the results of imaging studies--and an informatics geek.
Images are created on whatever imaging device--CT scanner, MR scanner, ultrasound machine, digital X-ray machine--and manipulated by the device's controlling system to do simple annotations, reformatting, etc. This is typically a Unix-based system running custom software designed and maintained by the device's vendor. The images are not usually interpreted on these systems.
From there, the images are sent to the PACS (Picutre Archiving and Communication System), which is just a gigantic central image database. These also tend to be Unix-based systems.
There tend to be two front-ends for looking at images in the PACS database. The first is the radiologist's interface, which is a high-end video workstation dedicated to showing medical images with the greatest possible fidelity. Most systems I've seen are Windows-based (Windows 2000, in our case) and run software which was built by the the imaging system vendors in the late 1990's. Much is made of the "lossless" nature of the images which are displayed; for example, when you log into such a machine, you're warned about how "This is a medical device" and that you shouldn't mess with it. Much is also made of "diagnostic-quality monitors" and high-end video cards to drive the monitors. This is an artifact from the early days of digital imaging interpretation in radiology, when there was a great deal of concern about whether the quality of the digital images would be adequate for us to figure out what was going on in Grandma's chest X-ray if we weren't looking at a piece of acetate. Most of these concerns have died away, as the differences in resolution and dynamic range turned out to be relatively minor and the added conveniences of being able to manipulate the images digitally turned out to be huge. For example, the new LCDs I seen being put on PACS workstations are off-the-shelf Dell 22-inchers, as far as I can tell.
Finally, there are "non-diagnostic" interfaces to the PACS images, which do tend to be web-based. These are so non-radiologist doctors can look at the images, too. Some are IE-based, and use an ActiveX control to display the images, and some use a Java applet. These are displayed with lossy compression (since someone might want to look at them from off-site via a VPN), and officially are not allowed to be used for interpretation. And in fact, I wouldn't want to; it's a lot harder to see subtle things on them than on a full-blown PACS workstation. Part of that is just the interface (it's hard to use those stupid ActiveX/applet things) and part of it is crummy/mis-configured monitors, but I suppose compression artifacts could also play a role.
So, to review: you go see your doctor, Dr. Smith, in her office, and she orders a chest X-ray for you because you're coughing and have a fever. You come to the hospital, and the nice technologist takes frontal and lateral view of your chest on the digital X-ray machine. He then goes back to the X-ray control room, and sees that the images are pretty good, and so he sticks your name on them, and a marker of the date/time and his name, and so on, and then sends them to the hospital's PACS system. I (the radiologist) am working at my PACS workstation, going through the long list of all of the CT scans, MR scans, and X-rays taken in the hospital. I get to your chest X-ray and look at it; I don't seen any sign of pneumonia, so I write a report (the subject of a whole different set of informatics) that basically says "Clear lungs" and that gets entered into your electronic medical record. Then, Dr. Smith back in her office can see your X-ray via her Web-based interface. If she wonders about something she sees, she can call me up and say, "What's that stuff at the left ape
Perl + Eclipse is coming... you can see the beginning at the EPIC project home page.
(I can't wait! Go, guys, go!)
I'm a libertarian (small "l" for me, please).
In my philosophy, the entire purpose of government is to protect individuals' rights. For example,
In my view, a "corporation" should have no rights of its own. (I'm not even that crazy about the idea of liability avoidance that's supposed to be the whole idea of incorporation, but that's another story.) So it goes without saying that the government should be in the business of preventing corporations from screwing individuals.
I'm not a huge extremist, but I do become disgusted with the extent to which the US government, far from protecting its citizens from mega-corporations, eagerly helps them exploit its citizens.
I don't think the government would find that sufficiently satisfying. Such a "war" does not require massive invasions of civil liberties or decreases in government transparency in the name of "national security" to fight.
I'm betting we'll stick with our current "war" on terrorism.
In my experience, people who study science are much more broadly educated than people who don't (which seems obvious). But physics majors I've known have non-science interests (frequently music, interestingly) at a much higher rate than English majors have science-related interests. (Actually, have you ever noticed that many non-scientists consider technical stuff to be beneath them?)
As for myself, I studied biology, physics, and electrical engineering as an undergrad; I'm getting an MD and a PhD in neuroscience. As an undergrad, I took classes in history, political science, writing, and American literature. By coincidence, I was also heavily concentrated as an undergrad in theatre arts, so I do happen to know a little Shakespeare. And I still maintain an interest in almost all of those things, and regularly read about them.
But I don't consider myself unusualy well-rounded among others in science I encounter; on the contrary, I'm frequently impressed by the other interesting things that they do.
Given that the author is David Wheeler, it seems pretty likely that you're right.
After reading the analysis, two things sprang out at me. The first is that a lot of the stuff on a Linux system is meant for development, rather than just using the system. The second is that lots of the stuff on the list clearly is "application" and not anyone's idea of an "operating system".
Specifically, in the top ten, we have:
Development Tools
Applications
(Also in the top 20 are libgcj, teTeX, postgresql, and xemacs. And we won't get into the issue of whether Mozilla (#2) should be considered part of the operating system.)
So my question is, what's the size of the non-development/non-application stuff? What's the size of the kernel plus the essential utilities (most of which are GNU, as RMS points out ad nauseum)?
I found something interesting in the preliminary injunction. The judge is drawing his conclusions:
Wow! There's a federal judge that seems to think the Constitution says that we should all be able to surf and e-mail anonymously! I'd like the Constitution to say that, but I don't think it does.
--Or, if you're trying to listen tomorrow: http://www.npr.org/ramfiles/fa/20010604.fa.ram.
Sorry for the blatant karma whoring.
...is a wonderful interviewer. She really seems to understand enough about the relevant issues to ask intelligent questions, and she has a knack for making her guests comfortable enough to reveal themselves. The show's format helps, too; plenty of time for each interview (20 or 30 minutes), and no morons calling in.
Should be a really worthwhile listen; probably one of the best interviews Linus will do on his book tour.
--
The same as in the great CueCat debate.
How is AOL going to make money?
AOL thinks there are lots of ways to make money on messaging. One of their executives is quoted as saying that Instant Messaging is AOL's most important asset.
But clearly, they think they can only succeed if they control the client as well as the servers. And a route into their system where they don't control the client is apparently unacceptable to them.
Their privacy policy is quite funny. It says, "Digital:Convergence wants every member of its community to understand what information we collect and how we intend to use that information."
But while the rest of the page is quite explicit about not giving out your phone number, it says nothing about the mother lode of data they (surely) collect about what you actually scan. In fact, they don't even mention the possibility!
...has nothing to do with IP or any other crap like that.
It seems pretty clear to me that this company is all about collecting personal information about you. Now, we're used to companies like Doubleclick et al. using various schemes to figure out where you browse, and generate models of your on-line behaviors automatically. But the CueCat thing is designed so that they can find out about things in the real world, too. They WANT you to scan things in your house (books, products, magazines, TV shows, etc), and when you do, their software does a quick lookup for you, and gives you the shovelware connected to whatever it is.
But it ALSO remembers who scanned what, and they can generate a much more impressive profile of you as a consumer: what do you watch, what magazines are you reading, what books do you have, etc. And I'll bet that THAT's what their business model is all about---more invasive consumer profiling.
And THAT's why they can't tolerate the Linux drivers which don't do the lookup through their servers---because they don't let them form that profile. They've probably calculated their (hugely expensive) giveaway scheme to give a particular rate of data generation through their servers, and if they don't get that, then their business model is shot to hell. Of COURSE they're going to try to prevent people from getting around their software: every person using the Linux driver is the same as a reader thrown away to them. So I'm sure that if a developer tried to work with them, they'd spend a lot of effort urging them to direct all connections through the Digital Convergance servers.
My father-in-law gave me his reader (which some magazine helpfully mailed him), and I was going to use it under Linux, but I've changed my mind. If I want a barcode reader, I'll go buy one that's free of this kind of crap.
It's another one of those prickly problems, I guess, if the law is going to have that little disregard for the idea of "responsibility". But in the end, it sounds like it's going to come down to some kind of monitoring scheme, right?
And maybe I'm hopelessly naive, but it seems like what you need to do is find a way to involve your employees in the process. Tell them what the problem is. Ask what kind of monitoring they'd find acceptable, and what they'd find intrusive. When you figure out what precisely your options are, ask your employees what they think of all of the different possibilities.
And when you figure something out, tell people as EXACTLY as you can what you're doing, and how it works. Tell them what you expect of them; tell them what limits you are going to respect (and then respect them, of course).
I'm sure some people will be a little put out, no matter what you come up with, but given that one can get all kinds of free e-mail accounts and that most have abundant access to the net, some restrictions on what one does at work shouldn't be too onerus.
What would it be illegal to stand on the border and shout into France?
Or, even more accurately, what would it be illegal to say over the telephone in France, regardless of where you are?
Or, yet more precisely, what would it be illegal for you to put as the outgoing message on your answering machine that anyone in France might call and hear?
Actually, in my first draft, the sentence ended, "... (except maybe Al Gore)." (No kidding. :)
I took it out because I thought it was needlessly inflammatory. Besides, maybe if we ignore those people, they'll go away. (I wish...)
But I don't think that's bad; in fact, it makes my inner libertarian quite happy. That's what makes the web so great! I can put anything I want on my site, and you can put anything you want on your site (including any URL you like).
And of course, it goes without saying that just because you ask me for a page I don't have to give it to you; I might put conditions on that. (Like charging you for it, or making you look at an ad or a silly splash screen.)
The only way there could be a legal recourse is if the government has the right to tell me how my site has to look and work, and I don't think anyone wants that.
Taxation on internet transactions is inevitable. We can argue over whether it's a good thing or not; an interesting contribution to the debate is Paul Krugman's oped piece in the NY Times (free registration, blah, blah, blah).
It seems to me, though, that the reason it isn't already taxed is not this garbage about strangling the new economy but the fact that it's the states that want to collect the taxes, and the Constitution specifically prohibits that. The politicians are just pretending to care about the Internet until they can figure out a way to tax it that's not against the constitution.
Hey! I want to use Mozilla with it!
Seriously, though, at least in the Red Hat RPM they distributed, they seem to have hard-coded in the assumption that you're using Netscape. In fact, when you install the RPM, it tries to start Netscape and connect to the Real site (even though one usually is root for installing the RPM). Not cool...
And there isn't apparently any place to indicate an alternate browser to be used...
I downloaded the player this morning, and installed it---it's quite sweet. Works great with ESD, sounds very good. I'm quite pleased with it.
But there's an odd trend I've noticed... When I went to listen to the live stream of my NPR station with my spiffy new player, it couldn't play it because it was for Windows Media Player only. Doh! I wrote them to tell them that this was pretty frustrating (especially since I'm pretty sure they also used to webcast for Real player).
But the fact is, this is becoming more and more common, I think. Sites are going to WMP-only based content. Is Microsoft trying to undercut their competition again by bundling software with the OS? (And is the DoJ paying attention?)
At least now, even non-hackers can get pseudo-privacy (which, while not absolute, is better than nothing). This proposal would make a bandit out of anyone who tries to be anonymous. And does anyone think that those who would engage in questionable practices will have much problem getting around whatever enforcement is put in place? (How does that silly NRA slogan go? "When you outlaw anonymity, only the outlaws will be anonymous.")
In response to your proposal (which is interesting), how would the anonymity it creates be fundamentally different from the illusion we have now?
--
Happy Premise #3: Even though I feel like I might ignite, I probably won't.
It seems to me that "meme" brings with it an interesting connotations:
:)
(1) Memes spread;
(2) Memes compete with one another;
(3) People study the nature of (1) and (2).
So it seems like it might be a useful shorthand for "an idea spreading in competition with others".
But it also makes us sound geekier.
Terry
It's a fairly bogus idea (in both senses, if you'll pardon the expression).
We're able to encode (and therefore to transmit and reconstruct) visual images digitally because we understand the relevant paramaters of digital images: intensity vs. two spatial dimensions and wavelength. Sound is even easier: intensity vs. time. It's even easier for these because to a reasonable approximation, these parametres are linear.
But we have NO idea of what the relevant parameters of a smell are. People have suggested all kinds of ideas over the years, but nothing has been really satisfactory. You've probably heard about the seven primary scents; that's pretty much a load of garbage, debunked by people who studied olfaction in the fifties and sixties. More recently, people have tried to understand the sense of smell in terms of the shapes of the molecules being detected, but even that has been unsatisfactory.
Perfumers, therefore, remain essentially what they were two hundred years ago: empiricists surrounded by a LOT of bottles. Hundreds of them. It's amazing to see. It takes YEARS to train someone to be good at it, and even then, their ability to reproduce an arbitrary odor is VERY limited. The notion that someone could put that in a box and sell it to people is absurd.
What you're much more likely to get is an odor equivalent of Jelly Belly jelly beans. You know how on the jelly beans, they say, "Combine cinnamon and popcorn to get the flavor of cinnammon popcorn balls"? Well, they'll choose a handful of cheap, fake scents, and use those over and over again. So you'll get things like "flowers" and "fruit" or even "fruity flowers", but never "the beach in Tahiti after a rainstorm".
Do YOU think people will pay for that?