The resolving power of a telescope is limited by the size of its aperture: the finite size of the aperture creates diffraction effects that effectively blur the image. The larger the aperture, the smaller these effects are, making it possible to resolve smaller objects. A larger aperture also increases the light-gathering power, making it possible to see fainter objects
The clever bit is that if you have two telescopes hooked up with interferometry, their effective aperture is equal to the distance between them. For radio telescopes, this has been used to create telescopes "as large as the Earth". It's a much harder thing to pull off in visible light wavelengths because of the precision required to align the telescopes.
This is how this sort of thing could work - I don't know if this is how Motorola did it, but it's how I'd expect it to be done.
Motorola forms a company, Iridium Inc., which they give some amount of money to. Iridium Inc. also gets money from other investors, who get an equity stake in it in return (i.e., they own part of the new company). It further borrows money from banks who are willing to lend it cash, and issues bonds which pay a certain amount of interest. The banks and the bond buyers are betting that its business will work, and that they'll get the interest and eventually get the principal back.
Iridium Inc. then spends all this money on satellites and launch costs. Nobody uses its phones and it runs out of money. The first thing that happens is that it stops paying interest on its bonds and loans. Even that isn't enough, and it declares bankruptcy and is dissolved (there are several possible layers of bankruptcy, but in this situation they blow through all of them). All its assets are sold off; this is how the Canadian company bought the satellites. The pittance that doing this brings in is distributed to all the people Iridium Inc. owes money to.
So who lost $4,975,000,000? Motorola lost the money it put in, as did all the investors, banks, and bond-holders. The risk was spread out, so it didn't wipe Motorola out. Because it was a separate company, its bankruptcy doesn't affect Motorola directly: Motorola isn't on the hook for the money everyone else lost. They knew what they were doing when they invested, and this sort of thing is a normal risk of investing in any business.
That should bring me up to the "screw up" question, so I will now relate a story that showcases my stupidity. One day I was working
on Slashdot's database, and cleaning up a lot of old database entries, when I suddenly realized I'd forgotten a "WHERE" clause in one
of the SQL statements.
ROLLBACK usually fixes these sorts of things, unless you've already gone past COMMIT, or unless you're using a database without transaction support.
Oh, wait, this isn't the latest MySQL vs PostgreSQL thread?
Something that's always struck me as odd about the DMCA is that something that cirvumvents an access control is illegal. This means that if I create some content, and "encrypt" it in some way (this needen't be real encryption as long as it prevents the content being viewed), anything that undoes that encryption is illegal, if it does it without my consent.
This implies that if I created a DVD encrypted with CSS, and my intention was that it should be viewed ONLY with DeCSS (hey, in a UCITA state I could probably even make that one of the terms and conditions of the DVD's license agreement), then playing it in a MPAA-approved DVD player would constitute circumventing its access control. So I should be able to prevent the MPAA from creating DVD players... After all, does it say in the DMCA "CSS belongs to MPAA and only they may use it as their access control method"? (OK, there may be patent issues. Suppose it's 17 years from now, the patents have expired, and the DMCA and MPAA are still around).
It gets worse: anything that circumvents the access control is illegal, and as the content creator, I get to choose my method of access control. Maybe I choose rot-17 - is tr now illegal? Maybe I choose zlib compression as my access control - is gzip now illegal?
That's a bit of a stretch and I'm sure no court would agree. It comes down to the interpretation of "effectively preventing access" to the content. If there are common tools to create content in that form, and common tools to view content in that form, then the access control is not "effective", in my opinion.
This makes me wonder if non-MPAA-approved CSS encryption tools would change how things work: bundle an encryptor with DeCSS and market it as a DVD mastering tool plus software decoder. Make a trivial change to the bitstream format (e.g., add a comment "Created by FooCSS") and make sure the DeCSS won't touch anything without that comment (presto, it's no longer an access control circumventer for MPAA's content). It would require a real stretch for this software bundle to be illegal under the DMCA, yet it nonetheless contains a functional CSS decryptor.
Nope - DOS was a rip-off of CP/M-86, which was a port to the 8086 of CP/M, a popular OS for 8080s and Z-80s. CP/M in my opinion was a rip-off of DEC's RT-11, which has its own lineage.
For example, the character used in RT-11 to specify options to commands was "/"; this was copied into CP/M and then into QDOS, which became DOS. And thus the slash/backslash wars began...
People are saying "the deregulators allowed wholesale prices to float but kept retail prices fixed" and thus blaming the deregulation (either for going too far or not going far enough). What they're missing is the fact that this deregulation deal was pushed for by THE POWER COMPANIES. The same companies that are now screaming that this is terrible, all the deregulator's fault, and must be fixed NOW were as pleased as punch when this deal was put together. They were betting that the wholesale price would drop under competition and they'd pocket a nice profit because they wouldn't have to lower retail prices.
They lost their bet, and now want their customers to eat the unexpected costs. So if wholesale prices fall, the customers lose, and if wholesale prices rise, the customers lose... sounds like a typical corporate sweetheart deal. They should not be allowed to get away with this sort of thing.
Oh, and the huge losses they're suffering are a lot lower than they claim - PG&E (for example) owes a lot of money to itself. (Well, to another wholly-owned subsidiary of the same parent company). Take that out of the equation and the cries of "Bail us out now or we declare bankruptcy and the lights go out" sound less like real panic and more like part of the negotiation for the next sweetheart deal at the expense of all Californians.
Furthermore, a simple vote procedure should be backed up by a strong computerized system, in order to ensure the rapid tabulation of results. Or would you have us go back to the time when it took weeks to figure out the election, every election?
The top priority in an election is to have an accurate, fair count (on top of fair, open voting procedures). A fast count is a nice thing, but it's certainly WAY down on the list of priorities. Is the election for the benefit of the public or the media?
Also, a fast count is not incompatible with pencil and paper voting. Remember, the Canadian general election held in November was counted quickly enough that the outcome was known before most people went to bed.
Another thing to note is that the Canadian election cycle, from announcement through campaigning to institution of the new government took SIX WEEKS. Compare that with the perpetual campaigning in the US, caused by the rigid election cycle - and remember that these drawn-out campaigns are a primary reason that politicians place so much emphasis on fund-raising, and thus contribute to the domination of the government by special interests.
The bilevel compression is done with SPM (Soft Pattern Matching - see my comment below). It uses a JBIG-like context model to do pixel coding, and uses an arithmetic coder called the Z-Prime coder, which is mathematically related to IBM's Q-coder.
Group 3 and Group 4 fax do really badly on halftoned images; JBIG-like compressors can do quite a bit better. JBIG2 includes a special mode to compress halftoned images.
TIFF6 doesn't support the MRC model used in DjVu - you have to go to TIFF-FX to get that.
Even for black and white images, PDF is the wrong way to go when you're dealing with scanned input. You either use their OCR which (like all OCR) introduces error, or you store the scanned image. The compression they use for the scanned image is quite poor. Modern compression methods like JBIG2 and the SPM compression used in DjVu can reduce a scanned document to a tiny fraction of its original size - 100:1 is not uncommon.
You're missing the advantage that MRC-based image formats like DjVu (and TIFF-FX and eventually JPEG-2000) have: by separating continuous-tone content like photographs from monotone (spot colour) high-frequency content like text, you can achieve really good compression while maintaining image quality. You don't want to do lossless compression of high-resolution photographic images - even with the best lossless compressor, the files are still huge. And if you do lossy compression of mixed content (text over images), you'll get really lousy results unless you separate the different types of image content.
The imaging model used by DjVu to get such good compression is to divide the image into (typically) three planes: a background plane (colour, compressed using wavelets, usually at a low resolution), a foreground plane (colour, compressed using wavelets, also at low resolution) and a selector plane. Imagine an ad page in a magazine: there's coloured text printed across a photographic image. The photograph is stored in the background, the text *colours* are stored in the foreground as blobs of colour (e.g., a blue word would have a large blue blob in the foreground), and the text *shapes* are stored in the selector. The decoder decompresses all three, and then draws the foreground colour where the selector is '1' and the background colour where the selector is '0'.
This scheme has the advantage that it keeps high-frequency content (the edges of the text) out of the background, enabling it to compress better (wavelets and JPEG don't handle high frequencies well: they either smear them, add ringing, or require a lot of bits). Another advantage is that there are specialised algorithms for compressing binary (1 and 0) images containing text. The one used by DjVu is called 'SPM' (Soft Pattern Matching). It works (roughly) by breaking the text image up into isolated characters, then deciding which characters look similar to each other. On most pages, there will be a lot of repeated characters. You can then get improved compression by storing only one instance of each unique character shape. SPM does something fancier than that but that's the basic idea.
This three-layer model is known as MRC (Mixed Raster Content). It's used outside DjVu - most notably in a file format called TIFF-FX. This is an extension of TIFF intended for Internet Fax (including colour fax). TIFF-FX is, if I remember right, RFC2301. There is an extension to TIFF-FX in the works to add support for JBIG2. What's JBIG2? It's a format standardised by ISO for doing bilevel image compression, using the same concept of improving text compression by identifying repeated shapes. (Actually, there's a lot more in JBIG2 but I'll leave that aside). The concepts used in DjVu's SPM were incorporated into the JBIG2 design (AT&T participated in the design of JBIG2), so JBIG2 can do anything SPM can do.
JBIG2 was approved by ISO and ITU this past spring.
Another file format that includes a lot of the same concepts as DjVu (MRC, repeated shape compression) is ScanSoft's XIFF - it's yet another TIFF extension, and is used in ScanSoft's Pagis line of products. In fact, much of the stuff in TIFF-FX is a standardised version of features that appeared in XIFF. MRC is also going to be in one of the later parts of JPEG-2000.
I've been involved in the work on XIFF and TIFF-FX and I was the editor of the JBIG2 standard. So what do I think of DjVu? One problem is that AT&T abandoned the standards process - they felt it was moving too slowly. This means that they opted for a proprietary solution over an open standard one. Yes, the standards process is slow, but when it works it produces things like JPEG and PNG (a W3C standard) - formats that everyone can use, and where you have a choice of encoders, a choice of decoders, and few worries about interoperability. Another problem is its reliance on arithmetic coding - while that gets you good compression, it can have speed problems. JBIG2 offers the choice of arithmetic compression (for applications where size is the most important issue) and Huffman-based compression, for applications where speed is the most important issue. I've personally seen a JBIG2 decoder decompress at over 1 gigapixel per second...
There are already unmanned surveillance aircraft; one such system is Predator. It can stay airborne up to 40 hours, and has been used in Yugoslavia.
There are some computer vision research projects whose goal is to provide automated analysis of Predator video (e.g., counting and identifying vehicles travelling on a road).
..these days, the land PARC's sitting on is worth more than the business unit.
If I remember right, Xerox doesn't own the land or the building - they did at one point, but sold it off to someone and leased it back. This was done so that the land and buildings would no longer be an asset, sitting on Xerox's financial books. Xerox (and a lot of other companies) measure success in terms of return on assets: what was our income as a percentage of the stuff we own? This kind of deal reduces the total assets, and increases ROA.
Basically it's a trick to improve a (short-term) measure of performance (ROA) but it has the side effect of decreasing actual (long-term) performance (you'll be paying that lease, at rates that will go up over time, forever). Sadly, this kind of trick is very common...
As long as they don't see a massive brain drain from the buyout, they should be doing fine. Despite what you say, there still exist companies like Microsoft and IBM who have enormous R&D departments and budgets. A buyout would be a shoe-in.
The brain drain is already under way - I know because I'm one of the drainees, having left PARC this past February to go to a startup. A number of other people left around the same time I did, and I've heard of a number more since then. Remember, PARC is located in the heart of Silicon Valley, meaning that you can't walk down the street without three companies trying to hire you - the sort of talent that PARC attracts is particularly in demand. The temptation to go to a hot company that is offering a high salary and large stock option package is quite high.
The exodus was bad enough in the spring, when Xerox's fortunes weren't quite so bad. With the company continuing to fumble, I can only imagine the morale there now (this is sad - the people there are friends; I worked with them for years).
Personally, I don't put a lot of credit in this rumour; for one thing, I don't see a clear buyer. Maybe HP or IBM, but they've both already got large Bay Area research facilities. Remember - the inventions PARC creates belong to Xerox, not to PARC itself, so what can you sell? All the patents? Xerox is using a lot of them. The buildings? Xerox doesn't own those. The employees and ongoing projects? That's possible, I suppose... there are a bunch of really cool projects going on there (including ones I worked on), and a lot of really bright people still there. They're working as hard as they can to create Xerox's future, but that's always been a long-term thing, and the short-term needs may trump that.
If Xerox does end up selling PARC, I don't know what will come out the other side - but it won't be the PARC that I worked at.
Re:I'm a Maths Graduate but ...
on
Does P = NP?
·
· Score: 3
Prof. Cook is a respected authority on complexity theory, so if he is doubtful about this paper, I'd take an extra large grain of salt with it...
Sufficiently respected that the first major theorem on NP-completeness (the proof that 3-SAT is NP-complete) is known as "Cook's theorem".
An aside: some friends who took their undergraduate computational complexity course from Prof. Cook said that he always referred to that theorem as "Theorem 3.71 in the book".
Re:Traveling Salesman
on
Does P = NP?
·
· Score: 2
Start in one corner and work in one basic direction. It gives you a path in P time, that is moderately close to optimal.
Unfortunately, moderately close to optimal isn't good enough. There's an entire field that's devoted to coming up with algorithms that are approximate solutions to NP-complete problems. These algorithms are usually expressed in terms of bounds: this algorithm is guaranteed to find a path that is no more than 2 times the length (or 1.5 times, or 1.1 times...) of the optimal path. So this kind of thing is definitely worth exploring, but an approximate solution to an NP-complete problem gets us no closer to knowing if P=NP.
If only the Supreme Court heard every case, we *might* stand a chance.
People should be aware that the upcoming Presidential election is likely to have a very great effect on the Supreme Court: 2-3 justices are likely to retire in the next four years, and many important cases recently have been decided by 5-4 votes. If you're not going to vote because you think that it won't matter to you which candidate gets elected, think of the secondary consequences like this. Ask yourself which candidate is more likely to appoint judges that will defend 1st amendment (and other) freedoms against corporate power grabs.
We need fewer justices like Scalia and Thomas, and more like Breyer and Souter.
With really high resolutions, you also run into a lot of other problems where fixed-size bitmaps were used: it becomes practically impossible to distinguish between all the tiny 16x16 toolbar icons in Windows applications, for example.
When I worked at PARC I had one of the prototype 7 megapixel displays which had a resolution up near 300dpi (282dpi if I remember right). It was 4-bit grayscale only, not colour, but text looked REALLY nice. A lot of Web sites really sucked, though, because they used frames or tables whose sizes were specified in terms of an absolute number of pixels, which usually meant that I'd see a column containing about 2-3 words per line, since each character on my display was 3x wider (in pixels) than they'd been expecting.
Antialiased text on this display was just beautiful.
MP3 is part of the MPEG standard, which was developed under the auspices of ISO. ISO has a patent policy, which basically states that anyone contributing to a standard must be willing to offer licenses to any of their intellectual property (usually patents) that is REQUIRED TO IMPLEMENT that standard. These licenses must be offered to all people requesting one, and must be for "reasonable terms and conditions". That doesn't mean free - the IPR owner can set their own fee schedule, as long as it's not outrageous.
The intent is that the requirement that anyone can get a license will stop companies from saying "we'll contribute our technology to a standard", then turning around and refusing to license it to their main competitor. That's OK as far as it goes. However, the fact that the holder can set any fees they want as long as they're "reasonable" is where the problems usually arise. When only big companies implemented standards, this wasn't a bid deal - what's $5K plus $0.10 per copy to a multinational? With the advent of open source, and decoders for standards being given away in boxes of cereal, any fee at all can make adoption of a standard go very slowly.
The ISO committees that put these standards together have a very tough job (I was the editor of an image compression standard, just approved last month - I know how much work goes into putting a standard together). A big part of that job is the desire to make sure that IPR hassles won't block widespread of adoption of the standard. JPEG is often used as an example of how well things can go - and it's also a good example of how poorly things can go. Not many people who haven't worked closely with JPEG know that the standard includes a lot of features that just aren't used. It's no coincidence that many of these unused features require getting patent licenses (not free ones). So the committees spend endless hours wrangling about intellectual property: attempting to get holders to make their licenses free, revising the draft standard if there's an IPR holder who is being intransigent, and so on. This isn't always entirely successful, but it usually gets most of the IPR holders lined up, reducing the problem and sometimes entirely eliminating it. In some cases (as with JPEG) a baseline profile is defined that avoids all IPR (if this is possible).
Slashdot Mirror
The clever bit is that if you have two telescopes hooked up with interferometry, their effective aperture is equal to the distance between them. For radio telescopes, this has been used to create telescopes "as large as the Earth". It's a much harder thing to pull off in visible light wavelengths because of the precision required to align the telescopes.
Motorola forms a company, Iridium Inc., which they give some amount of money to. Iridium Inc. also gets money from other investors, who get an equity stake in it in return (i.e., they own part of the new company). It further borrows money from banks who are willing to lend it cash, and issues bonds which pay a certain amount of interest. The banks and the bond buyers are betting that its business will work, and that they'll get the interest and eventually get the principal back.
Iridium Inc. then spends all this money on satellites and launch costs. Nobody uses its phones and it runs out of money. The first thing that happens is that it stops paying interest on its bonds and loans. Even that isn't enough, and it declares bankruptcy and is dissolved (there are several possible layers of bankruptcy, but in this situation they blow through all of them). All its assets are sold off; this is how the Canadian company bought the satellites. The pittance that doing this brings in is distributed to all the people Iridium Inc. owes money to.
So who lost $4,975,000,000? Motorola lost the money it put in, as did all the investors, banks, and bond-holders. The risk was spread out, so it didn't wipe Motorola out. Because it was a separate company, its bankruptcy doesn't affect Motorola directly: Motorola isn't on the hook for the money everyone else lost. They knew what they were doing when they invested, and this sort of thing is a normal risk of investing in any business.
Nah - did you know that "< file" can appear anywhere on the command line - including before the command!
/etc/nsswitch.conf wc
$ <
32 130 936
$
So to insert your frob filter doesn't involve any cutting and pasting.
Oh, wait, this isn't the latest MySQL vs PostgreSQL thread?
This implies that if I created a DVD encrypted with CSS, and my intention was that it should be viewed ONLY with DeCSS (hey, in a UCITA state I could probably even make that one of the terms and conditions of the DVD's license agreement), then playing it in a MPAA-approved DVD player would constitute circumventing its access control. So I should be able to prevent the MPAA from creating DVD players... After all, does it say in the DMCA "CSS belongs to MPAA and only they may use it as their access control method"? (OK, there may be patent issues. Suppose it's 17 years from now, the patents have expired, and the DMCA and MPAA are still around).
It gets worse: anything that circumvents the access control is illegal, and as the content creator, I get to choose my method of access control. Maybe I choose rot-17 - is tr now illegal? Maybe I choose zlib compression as my access control - is gzip now illegal?
That's a bit of a stretch and I'm sure no court would agree. It comes down to the interpretation of "effectively preventing access" to the content. If there are common tools to create content in that form, and common tools to view content in that form, then the access control is not "effective", in my opinion.
This makes me wonder if non-MPAA-approved CSS encryption tools would change how things work: bundle an encryptor with DeCSS and market it as a DVD mastering tool plus software decoder. Make a trivial change to the bitstream format (e.g., add a comment "Created by FooCSS") and make sure the DeCSS won't touch anything without that comment (presto, it's no longer an access control circumventer for MPAA's content). It would require a real stretch for this software bundle to be illegal under the DMCA, yet it nonetheless contains a functional CSS decryptor.
They lost their bet, and now want their customers to eat the unexpected costs. So if wholesale prices fall, the customers lose, and if wholesale prices rise, the customers lose... sounds like a typical corporate sweetheart deal. They should not be allowed to get away with this sort of thing.
Oh, and the huge losses they're suffering are a lot lower than they claim - PG&E (for example) owes a lot of money to itself. (Well, to another wholly-owned subsidiary of the same parent company). Take that out of the equation and the cries of "Bail us out now or we declare bankruptcy and the lights go out" sound less like real panic and more like part of the negotiation for the next sweetheart deal at the expense of all Californians.
Also, a fast count is not incompatible with pencil and paper voting. Remember, the Canadian general election held in November was counted quickly enough that the outcome was known before most people went to bed.
Another thing to note is that the Canadian election cycle, from announcement through campaigning to institution of the new government took SIX WEEKS. Compare that with the perpetual campaigning in the US, caused by the rigid election cycle - and remember that these drawn-out campaigns are a primary reason that politicians place so much emphasis on fund-raising, and thus contribute to the domination of the government by special interests.
Group 3 and Group 4 fax do really badly on halftoned images; JBIG-like compressors can do quite a bit better. JBIG2 includes a special mode to compress halftoned images.
TIFF6 doesn't support the MRC model used in DjVu - you have to go to TIFF-FX to get that.
Even for black and white images, PDF is the wrong way to go when you're dealing with scanned input. You either use their OCR which (like all OCR) introduces error, or you store the scanned image. The compression they use for the scanned image is quite poor. Modern compression methods like JBIG2 and the SPM compression used in DjVu can reduce a scanned document to a tiny fraction of its original size - 100:1 is not uncommon.
You're missing the advantage that MRC-based image formats like DjVu (and TIFF-FX and eventually JPEG-2000) have: by separating continuous-tone content like photographs from monotone (spot colour) high-frequency content like text, you can achieve really good compression while maintaining image quality. You don't want to do lossless compression of high-resolution photographic images - even with the best lossless compressor, the files are still huge. And if you do lossy compression of mixed content (text over images), you'll get really lousy results unless you separate the different types of image content.
This scheme has the advantage that it keeps high-frequency content (the edges of the text) out of the background, enabling it to compress better (wavelets and JPEG don't handle high frequencies well: they either smear them, add ringing, or require a lot of bits). Another advantage is that there are specialised algorithms for compressing binary (1 and 0) images containing text. The one used by DjVu is called 'SPM' (Soft Pattern Matching). It works (roughly) by breaking the text image up into isolated characters, then deciding which characters look similar to each other. On most pages, there will be a lot of repeated characters. You can then get improved compression by storing only one instance of each unique character shape. SPM does something fancier than that but that's the basic idea.
This three-layer model is known as MRC (Mixed Raster Content). It's used outside DjVu - most notably in a file format called TIFF-FX. This is an extension of TIFF intended for Internet Fax (including colour fax). TIFF-FX is, if I remember right, RFC2301. There is an extension to TIFF-FX in the works to add support for JBIG2. What's JBIG2? It's a format standardised by ISO for doing bilevel image compression, using the same concept of improving text compression by identifying repeated shapes. (Actually, there's a lot more in JBIG2 but I'll leave that aside). The concepts used in DjVu's SPM were incorporated into the JBIG2 design (AT&T participated in the design of JBIG2), so JBIG2 can do anything SPM can do. JBIG2 was approved by ISO and ITU this past spring.
Another file format that includes a lot of the same concepts as DjVu (MRC, repeated shape compression) is ScanSoft's XIFF - it's yet another TIFF extension, and is used in ScanSoft's Pagis line of products. In fact, much of the stuff in TIFF-FX is a standardised version of features that appeared in XIFF. MRC is also going to be in one of the later parts of JPEG-2000.
I've been involved in the work on XIFF and TIFF-FX and I was the editor of the JBIG2 standard. So what do I think of DjVu? One problem is that AT&T abandoned the standards process - they felt it was moving too slowly. This means that they opted for a proprietary solution over an open standard one. Yes, the standards process is slow, but when it works it produces things like JPEG and PNG (a W3C standard) - formats that everyone can use, and where you have a choice of encoders, a choice of decoders, and few worries about interoperability. Another problem is its reliance on arithmetic coding - while that gets you good compression, it can have speed problems. JBIG2 offers the choice of arithmetic compression (for applications where size is the most important issue) and Huffman-based compression, for applications where speed is the most important issue. I've personally seen a JBIG2 decoder decompress at over 1 gigapixel per second...
There are already unmanned surveillance aircraft; one such system is Predator. It can stay airborne up to 40 hours, and has been used in Yugoslavia. There are some computer vision research projects whose goal is to provide automated analysis of Predator video (e.g., counting and identifying vehicles travelling on a road).
Basically it's a trick to improve a (short-term) measure of performance (ROA) but it has the side effect of decreasing actual (long-term) performance (you'll be paying that lease, at rates that will go up over time, forever). Sadly, this kind of trick is very common...
The exodus was bad enough in the spring, when Xerox's fortunes weren't quite so bad. With the company continuing to fumble, I can only imagine the morale there now (this is sad - the people there are friends; I worked with them for years).
Personally, I don't put a lot of credit in this rumour; for one thing, I don't see a clear buyer. Maybe HP or IBM, but they've both already got large Bay Area research facilities. Remember - the inventions PARC creates belong to Xerox, not to PARC itself, so what can you sell? All the patents? Xerox is using a lot of them. The buildings? Xerox doesn't own those. The employees and ongoing projects? That's possible, I suppose... there are a bunch of really cool projects going on there (including ones I worked on), and a lot of really bright people still there. They're working as hard as they can to create Xerox's future, but that's always been a long-term thing, and the short-term needs may trump that.
If Xerox does end up selling PARC, I don't know what will come out the other side - but it won't be the PARC that I worked at.
People should be aware that the upcoming Presidential election is likely to have a very great effect on the Supreme Court: 2-3 justices are likely to retire in the next four years, and many important cases recently have been decided by 5-4 votes. If you're not going to vote because you think that it won't matter to you which candidate gets elected, think of the secondary consequences like this. Ask yourself which candidate is more likely to appoint judges that will defend 1st amendment (and other) freedoms against corporate power grabs.
We need fewer justices like Scalia and Thomas, and more like Breyer and Souter.
With really high resolutions, you also run into a lot of other problems where fixed-size bitmaps were used: it becomes practically impossible to distinguish between all the tiny 16x16 toolbar icons in Windows applications, for example.
When I worked at PARC I had one of the prototype 7 megapixel displays which had a resolution up near 300dpi (282dpi if I remember right). It was 4-bit grayscale only, not colour, but text looked REALLY nice. A lot of Web sites really sucked, though, because they used frames or tables whose sizes were specified in terms of an absolute number of pixels, which usually meant that I'd see a column containing about 2-3 words per line, since each character on my display was 3x wider (in pixels) than they'd been expecting.
Antialiased text on this display was just beautiful.
MP3 is part of the MPEG standard, which was developed under the auspices of ISO. ISO has a patent policy, which basically states that anyone contributing to a standard must be willing to offer licenses to any of their intellectual property (usually patents) that is REQUIRED TO IMPLEMENT that standard. These licenses must be offered to all people requesting one, and must be for "reasonable terms and conditions". That doesn't mean free - the IPR owner can set their own fee schedule, as long as it's not outrageous.
The intent is that the requirement that anyone can get a license will stop companies from saying "we'll contribute our technology to a standard", then turning around and refusing to license it to their main competitor. That's OK as far as it goes. However, the fact that the holder can set any fees they want as long as they're "reasonable" is where the problems usually arise. When only big companies implemented standards, this wasn't a bid deal - what's $5K plus $0.10 per copy to a multinational? With the advent of open source, and decoders for standards being given away in boxes of cereal, any fee at all can make adoption of a standard go very slowly.
The ISO committees that put these standards together have a very tough job (I was the editor of an image compression standard, just approved last month - I know how much work goes into putting a standard together). A big part of that job is the desire to make sure that IPR hassles won't block widespread of adoption of the standard. JPEG is often used as an example of how well things can go - and it's also a good example of how poorly things can go. Not many people who haven't worked closely with JPEG know that the standard includes a lot of features that just aren't used. It's no coincidence that many of these unused features require getting patent licenses (not free ones). So the committees spend endless hours wrangling about intellectual property: attempting to get holders to make their licenses free, revising the draft standard if there's an IPR holder who is being intransigent, and so on. This isn't always entirely successful, but it usually gets most of the IPR holders lined up, reducing the problem and sometimes entirely eliminating it. In some cases (as with JPEG) a baseline profile is defined that avoids all IPR (if this is possible).