Not only that, but suppose I have Internet access but no fax machine, and you have a fax machine and no Internet access. In the current situation, we simply can't exchange documents short of shipping dead trees to each other. The wider scope of internet fax encompasses gateways on both ends, so I could send a document to a gateway that will fax it to you, and you could fax a reply to a gateway that would send it on to me.
This kind of IP license agreement is very common in standards. In particular, that Xerox rights agreement is very similar to an IBM license agreement I recall from a different, but related standard.
The gist of these license agreements is: if we give away our IP for you to use, you have to give us your IP to use. This prevents the situation where company A puts a lot of effort into creating a standard (and it is a lot of effort - multiple employees working on it full-time, lots of travel expenses), and company B has some IP that is required to use it. A offers its IP under a license like this to prevent B from saying "hey, nice standard - thanks for the work, and start handing us the bucks".
Yes, this kind of grantback license only really protects A from B - everyone else can still get screwed by B. Some form of grantbacks say "you can use our patents in this standard for free, but only if you let everyone (not just us) use them for free" - I've seen that used too.
And don't forget: the effect of this license on you if you DON'T own any patents is nil: in return for something (the right to use Xerox's IPR), you in return are required to grant them: absolutely nothing (i.e., you own nothing so you owe them nothing).
The compression used in TIFF-FX is very different from the simple zlib-based compression in PNG: it is able to represent multi-layer documents, and use different compression methods on each layer.
Consider amagazine page with some coloured text over a photographic background. (Yes, internet fax isn't just black and white.) Using TIFF-FX, you can represent
the background photograph as JPEG
the shape of the text characters with JBIG2 (a compression standard that is optimised for compressing text)
the colours of the text characters using another compression method suited for that
This allows you to get MUCH higher compression than something like PNG that doesn't take into account the document structure.
Something I haven't seen mentioned in the discussion of this problem: in the article where the Russians describe the SQL Server 6.5 problem, they said that the frequency of the bug varied depending on the processor it was running on: below a certain processor speed (200MHz?) or above another speed (500MHz?) it happened more often; in between it happened less often.
To me, this just screams "race condition": two threads weren't synchronising correctly with each other. This not only explains why it wouldn't happen every time (sometimes the threads interact properly, sometimes one clobbers the other), but the variability due to processor speed. Race conditions can be very subtle bugs, and not at all evident even on close inspection of the code. They tend to show up more under load.
And when there's a fire, or an earthquake, or a lightning strike, you'll feel really dumb.
Backups involve more than just having a redundant copy of the data in case of a drive (or a few drives) failing. Offsite storage and archiving are just as important - and while you could do this with the "build a second array for backup" approach, it's not really practical to move these heavy boxes around on a daily or weekly basis.
Re:When will businesses be clueful?
on
Books on Demand
·
· Score: 1
Who the hell is going to transmit already-rasterised pages?! That would limit even DVDs to 470 pages. I don't even want to *think* about transmission times.
Just use the right compression method. I've achieved 1000:1 lossless compression on rasterised book pages - see my comment above. That was at 600dpi, and at 2400dpi I'd expect 4000:1 compression (the compression efficiency of JBIG2 increases with the resolution).
Actually, book publishers and printers are quite wary even of PostScript, and would have serious issues with DVI as well. You want each printing of a book to come out exactly the same as the previous one - no matter where it was printed, no matter which of the multiple machines at a site it was printed on. If the master electronic form for your book was a PostScript or PDF file, you can run into trouble: if one printer has a slightly different set of fonts loaded than another one, you can get font substitution, which you really don't want. Even if you control that problem, different PostScript interpreters actually produce slightly different output - the character spacing can differ visibly. There are other issues like halftone rendering (spacing, dot shape) that can lead to different appearances in the final paper document. Let's not even think about the differences in rendering between two different versions of Word, if you were insane enough to have your book master be a.DOC file.
One solution is to create a pre-master document (in PostScript, DVI, Word, or whatever), RIP it to a set of page bitmaps, proof the pages, and then use those page images from then on - you know that each printer will produce exactly the same bitmap. Page images are big, though, and if you want to keep a lot of books ready for print-on-demand, storage (or bandwidth to a central site) is an issue. In a previous job, I helped develop JBIG2, which is a compression format for black-and-white (1 bit per pixel) document images, which makes the space issue pretty much go away - we had a test document which was a 500-page book, rendered at 600dpi. The uncompressed page bitmaps were about 1.5GB, but we were able to achieve 1000:1 lossless compression using JBIG2. The entire book could fit on a floppy...
Rendering speed is also an issue - PostScript can be quite slow to interpret. One of JBIG2's design goals was very fast decompression, and we were able to print that 500 page book in under 3 minutes (on a Xerox 180ppm printer); the decompression time was well under 1 minute.
A couple of years ago, one of the international standards organisations (I don't remember which) got annoyed at the "mega=10^6" versus "mega=2^20" confusion, and came up with new prefixes. According to these:
1000 bytes = 1 kilobyte
1024 bytes = 1 kibibyte
1000000 bytes = 1 megabyte
1048576 bytes = 1 mebibyte
and so on - gibibytes, tebibytes, pebibytes, ebibytes, each larger than the previous one by a factor of 1024.
Even though the prefix names sound silly, I Am Not Making This Up.
What has the "copyright/patent/intellectual property" world contributed to networking? NETBEUI? Appletalk? IPX? Where are they now? Could the
Microsofts of the world ever "innovate" anything like TCP/IP without endless copyright/patent litigation and an IRS-like licensing scheme?
Ethernet.
It was developed by a for-profit corporation and patented. However, the corporation decided that making it a standard was more useful than keeping it proprietary, and as part of the standardisation process, the patents were licensed to all comers for a nominal fee (I think it was a few hundred dollars). Since the licensors were all hardware manufacturers, this was not much of a burden - if you're shipping a physical device, money is changing hands somewhere along the way so a nominal fee for patents isn't a barrier to using the standard. In the software world, of course, even a small fee can severely limit the use of a standard.
If you want to measure Microsoft's contribution to the greater good, look for standards that they have originated, where they gave up patent rights in the process. I mean standards in the sense of formal open specifications, where the specifications are under the control of an independent body, not de-facto standards that you have to reverse-engineer to use. I vaguely remember there being some MS-originated standards that meet this criterion - maybe DHCP?
MINORS CANNOT ENTER INTO BINGING CONTRACTS. Actually
they can, but they can also say "I want out" whenever they feel like.
"Daddy, I signed a contract with Timmy next door that we'd both drink three six-packs of Coors."
"No problem, Johnny - you can say 'I want out' whenever you feel like!"
Don't tell me that when you send **FOR FREE** a manuscript to a publisher, you immediately forfeit all copyright to the article to the
publisher????
Pretty much. All the journals and conferences I've dealt with require you to sign a copyright form assigning them some kind of exclusive rights, including in some cases online reprint rights.
In practice, this is subject to negotiation (e.g., my old employer required me to sign and send in their own copyright form, not the publisher's one), and the journals sometimes aren't very good about checking that a form even came back with the camera-ready copy. Also, some publishers have more liberal copyright agreements - I seem to recall that the IEEE allowed the author to provide online reprints of the articles in their journals, as long as proper credit was given to the IEEE on the Web page containing the reprint.
Peer review is generally anonymous: the editor knows who did the reviewing, but the author does not. Having the reviewers sign the article goes against this long-standing tradition. Sure, they could use pseudonymous signatures, but then you can tell that person X reviewed both this article and this other one - more information than is revealed currently.
Mind you, it's sometimes quite possible to have a good guess at who the reviewers are: after all, if the editor is doing their job, the reviewers will be experts in your particular sub-field, and so people you're likely to have met. Dead giveaways are things like a review comment saying "the following articles need to be added to the bibliography" followed by a list of articles all sharing an author.
In some cases, the identity of the author of the paper is hidden from the reviewers. It's usually not too hard to come up with a reasonable guess in this case.
The purpose behind this anonymity (both sides) is to attempt to improve the quality of the reviewing process: if you don't know who wrote a paper, you won't be tempted to give it a good review because the author is a buddy of yours coming up for tenure, or conversely give it a bad review because you've got a feud going on with the author. Keeping the reviewers' names secret is a way to try to prevent those feuds from starting in the first place.
The military's vision calls for a bomber that would be too fast to shoot down.
Maybe impossible to hit with a missile - but all you need to do with something going this fast is to fill the airspace in front of it with pebbles. Saturating a region is much easier than hitting a moving target. Saturating the right region isn't trivial, but I'd much rather figure out how to do that than how to build a Mach 7+ interceptor missile.
There's a little more to it than that: it's an inconspicuous error-tolerant high-density barcode. Inconspicuous means that you can embed data in what looks like just a background gray stipple. Error-tolerant means that you can scan it at low resolutions, at any angle, you can put it through lousy copying or fax it, you can even cut bits of it off or obscure them, and at the end you can still get the original data back. There are limits, of course, but the glyphs are pretty robust. At one point, one team working with DataGlyphs got a T-shirt made up with a glyph block printed on it. The individual marks were, of course, much larger than they would be on paper, and the glyph block had a number of other graphical elements (drawings etc) splattered on top of it. One of the team members, for grins, put the T-shirt on a scanner and fed the result into the glyph decoder - and despite the scale being way different from normal, and the missing glyphs (obscured by the stuff drawn over them), it correctly read back the encoded message.
Also, DataGlyphs are a couple of orders of magnitude higher density than standard barcodes. Two orders of magnitude is enough to make a qualitative difference in how something is used: applications that could not even be considered before are now easy, and so the way in which something is used can change drastically.
I never worked on DataGlyphs, but I was at PARC for some of the time that they were being developed. They've been around for quite a while and are actually in use in the real world. Those of you with insurance from State Farm probably get newsletters; each newsletter comes with a tear-off card that says "if you're not interested in receiving this newsletter, just drop this card in the mail". If you look at that card, there's what looks like a gray rectangle, but if you look closely you can see this is made up of a bunch of tiny '/' and '\' marks. That's a DataGlyph block. When State Farm gets one of those cards back, they just drop it onto a scanner, the glyph gets read out, and you get unsubscribed.
I happened to be standing around when the inventor of DataGlyphs came in with his copy of the State Farm newsletter - he seemed quite pleased to be receiving mail using his invention.
In cases like this, where open standards are involved, you can fight the power. The people on these committees aren't appointed by God - they get sent there by their companies to work for their companies' interests. The committee member ship is generall open - anyone who is willing to come to the meetings gets a vote.
This is based on my experience with the NCITS L3.2 committee, which is responsible for the US side of standards like JPEG. Its sister committee, L3.1, is responsible for the US side of MPEG. If T13 (the committee where CPRM was proposed) works anything like L3.2, then membership is open to any organisation willing to pay the membership dues (hundreds of dollars per year) and come to the meetings (three per year, scattered around the US). In order to become a voting member, your company or organisation has to have paid its dues and show up at at least two consecutive meetings. Each organisation gets one vote, even if they have several representatives in attendance.
It's not cheap - dues plus attending the meetings will take several thousand dollars per year, plus three weeks of work. It's mostly boring work, discussing tiny details of standards, making nit-picking changes to the wording, and so on. But don't complain that you're shut out - it's quite possible for a few people to pool together and form their own organisation and get just as much voting power as Apple, IBM, Microsoft, and so on. Three people sharing the dues and each going to one meeting per year wouldn't be a huge drain on money or time.
For the organisations like L3.2 and L3.1 that send delegates to international groups like JPEG and MPEG, you can also become a delegate. This is done on an individual, not organisation, basis: the individual delegate must have attended two out of the last three NCITS meetings (plus their organisation must be a member in good standing). Then you can attend three more international meeting a year, doing even more boring work, but often in interesting destinations, though as far as I could tell, all of the places I went looked like the inside of a conference room.
So it is a commitment of time and money, but anyone who wants to put in the effort can have a say on what standards are formed.
The people who signed the AUP and then spammed owe you $500. The credit card companies won't collect this debt for you (even though it seems they should). So take the spammers to small claims court - you have a signed agreement and proof they violated it - just because the CC companies won't enforce it doesn't mean it's not a real debt, and small claims is designed for collecting on debts like this. No lawyers needed.
I don't have depth perception: my eyes point in different directions and the images aren't fused. This means that any "3D" display technology that depends on stereo fusion just doesn't work. About 10% of the population has the same sort of problem as me. The Actuality display is real 3D, not cheating like stereo - you can move your head around to see the backs of things.
With the patient inside an MRI machine, they'd have to use unconventional surgical instruments: anything metallic would be really hard to control in the magnetic field. Scalpels made of obsidian?
Absolutely - small companies are more likely to grow manyfold than big companies, but they're more likely to go under too. But it only takes one huge winner to make up for lots of losers. This is how venture capital works: the investors are expecting 90% or more of the companies they invest in to fold - but if the 10% that don't each grow 20x (VCs get in early), they've got a 100% return on their investment.
Playing this kind of odds game is definitely risky. Once the money is invested, treat it as gone and then you'll be pleasantly surprised if you get anything at all back.
Clearly, you should invest in Red Hat - it's an awful lot easier to grow income of $27 million by a factor of 10 than it is to grow earnings of $2.58 billion by a factor of 10.
Rapid growth happens in small companies, not large ones, and that's where the greatest returns are to be found.
I just installed an SMC 802.11b/firewall combination (SMC7004WBR). You plug the WAN device (DSL in my case) in one port, internal network devices in another, and let wireless devices connect to it. It's a nice little box and seems to be working just great so far. However, the 802.11b network is on the inside of the firewall, which probably isn't what you had in mind.
On the other hand, it does have some access controls built in, so you can control what IP addresses (and ports, I think) different internal IP addresses have access to. It might be possible to configure this in such a way that IP addresses handed out via DHCP are prohibited from talking to anything else on the internal network. This would be a stretch since those access controls were designed for things like "don't let Junior's computer connect to FTP", so might not be up to the task of protecting different machines "inside" the firewall from each other.
Do you even KNOW what a PPC processor is? It is no great leap to go from 68k to PPC.. their cores are very similar, as are the compilers.
Let's see - a 2-operand 16-register model (split into 8 address and 8 data registers) with complex addressing modes (have you ever looked at the insane number of possible addressing modes in a 68020?) versus a 3-operand 32-register model with few addressing modes. I'd say that's quite a difference.
Not only that, but porting a WELL DESIGNED OS is not a terribly difficult task. Just because Microsoft is incapable of doing it doesn't make Apple that wonderful.
Porting a well-designed OS isn't too bad - but interestingly enough, that's not what Apple did. They used an interpreter to execute all the 680x0 code, including large parts of the OS. A lot of the OS was still 680x0 code until quite recently.
Making this work is quite a feat, but don't confuse it with good OS design.
Slashdot Mirror
Not only that, but suppose I have Internet access but no fax machine, and you have a fax machine and no Internet access. In the current situation, we simply can't exchange documents short of shipping dead trees to each other. The wider scope of internet fax encompasses gateways on both ends, so I could send a document to a gateway that will fax it to you, and you could fax a reply to a gateway that would send it on to me.
The gist of these license agreements is: if we give away our IP for you to use, you have to give us your IP to use. This prevents the situation where company A puts a lot of effort into creating a standard (and it is a lot of effort - multiple employees working on it full-time, lots of travel expenses), and company B has some IP that is required to use it. A offers its IP under a license like this to prevent B from saying "hey, nice standard - thanks for the work, and start handing us the bucks".
Yes, this kind of grantback license only really protects A from B - everyone else can still get screwed by B. Some form of grantbacks say "you can use our patents in this standard for free, but only if you let everyone (not just us) use them for free" - I've seen that used too.
And don't forget: the effect of this license on you if you DON'T own any patents is nil: in return for something (the right to use Xerox's IPR), you in return are required to grant them: absolutely nothing (i.e., you own nothing so you owe them nothing).
Consider amagazine page with some coloured text over a photographic background. (Yes, internet fax isn't just black and white.) Using TIFF-FX, you can represent
- the background photograph as JPEG
- the shape of the text characters with JBIG2 (a compression standard that is optimised for compressing text)
- the colours of the text characters using another compression method suited for that
This allows you to get MUCH higher compression than something like PNG that doesn't take into account the document structure.To me, this just screams "race condition": two threads weren't synchronising correctly with each other. This not only explains why it wouldn't happen every time (sometimes the threads interact properly, sometimes one clobbers the other), but the variability due to processor speed. Race conditions can be very subtle bugs, and not at all evident even on close inspection of the code. They tend to show up more under load.
Backups involve more than just having a redundant copy of the data in case of a drive (or a few drives) failing. Offsite storage and archiving are just as important - and while you could do this with the "build a second array for backup" approach, it's not really practical to move these heavy boxes around on a daily or weekly basis.
One solution is to create a pre-master document (in PostScript, DVI, Word, or whatever), RIP it to a set of page bitmaps, proof the pages, and then use those page images from then on - you know that each printer will produce exactly the same bitmap. Page images are big, though, and if you want to keep a lot of books ready for print-on-demand, storage (or bandwidth to a central site) is an issue. In a previous job, I helped develop JBIG2, which is a compression format for black-and-white (1 bit per pixel) document images, which makes the space issue pretty much go away - we had a test document which was a 500-page book, rendered at 600dpi. The uncompressed page bitmaps were about 1.5GB, but we were able to achieve 1000:1 lossless compression using JBIG2. The entire book could fit on a floppy...
Rendering speed is also an issue - PostScript can be quite slow to interpret. One of JBIG2's design goals was very fast decompression, and we were able to print that 500 page book in under 3 minutes (on a Xerox 180ppm printer); the decompression time was well under 1 minute.
Yeah, but "The sheep, a proud and noble animal" just doesn't have the same ring.
1000 bytes = 1 kilobyte
1024 bytes = 1 kibibyte
1000000 bytes = 1 megabyte
1048576 bytes = 1 mebibyte
and so on - gibibytes, tebibytes, pebibytes, ebibytes, each larger than the previous one by a factor of 1024.
Even though the prefix names sound silly, I Am Not Making This Up.
It was developed by a for-profit corporation and patented. However, the corporation decided that making it a standard was more useful than keeping it proprietary, and as part of the standardisation process, the patents were licensed to all comers for a nominal fee (I think it was a few hundred dollars). Since the licensors were all hardware manufacturers, this was not much of a burden - if you're shipping a physical device, money is changing hands somewhere along the way so a nominal fee for patents isn't a barrier to using the standard. In the software world, of course, even a small fee can severely limit the use of a standard.
If you want to measure Microsoft's contribution to the greater good, look for standards that they have originated, where they gave up patent rights in the process. I mean standards in the sense of formal open specifications, where the specifications are under the control of an independent body, not de-facto standards that you have to reverse-engineer to use. I vaguely remember there being some MS-originated standards that meet this criterion - maybe DHCP?
"No problem, Johnny - you can say 'I want out' whenever you feel like!"
Mind you, it's sometimes quite possible to have a good guess at who the reviewers are: after all, if the editor is doing their job, the reviewers will be experts in your particular sub-field, and so people you're likely to have met. Dead giveaways are things like a review comment saying "the following articles need to be added to the bibliography" followed by a list of articles all sharing an author.
In some cases, the identity of the author of the paper is hidden from the reviewers. It's usually not too hard to come up with a reasonable guess in this case.
The purpose behind this anonymity (both sides) is to attempt to improve the quality of the reviewing process: if you don't know who wrote a paper, you won't be tempted to give it a good review because the author is a buddy of yours coming up for tenure, or conversely give it a bad review because you've got a feud going on with the author. Keeping the reviewers' names secret is a way to try to prevent those feuds from starting in the first place.
Also, DataGlyphs are a couple of orders of magnitude higher density than standard barcodes. Two orders of magnitude is enough to make a qualitative difference in how something is used: applications that could not even be considered before are now easy, and so the way in which something is used can change drastically.
I happened to be standing around when the inventor of DataGlyphs came in with his copy of the State Farm newsletter - he seemed quite pleased to be receiving mail using his invention.
This is based on my experience with the NCITS L3.2 committee, which is responsible for the US side of standards like JPEG. Its sister committee, L3.1, is responsible for the US side of MPEG. If T13 (the committee where CPRM was proposed) works anything like L3.2, then membership is open to any organisation willing to pay the membership dues (hundreds of dollars per year) and come to the meetings (three per year, scattered around the US). In order to become a voting member, your company or organisation has to have paid its dues and show up at at least two consecutive meetings. Each organisation gets one vote, even if they have several representatives in attendance.
It's not cheap - dues plus attending the meetings will take several thousand dollars per year, plus three weeks of work. It's mostly boring work, discussing tiny details of standards, making nit-picking changes to the wording, and so on. But don't complain that you're shut out - it's quite possible for a few people to pool together and form their own organisation and get just as much voting power as Apple, IBM, Microsoft, and so on. Three people sharing the dues and each going to one meeting per year wouldn't be a huge drain on money or time.
For the organisations like L3.2 and L3.1 that send delegates to international groups like JPEG and MPEG, you can also become a delegate. This is done on an individual, not organisation, basis: the individual delegate must have attended two out of the last three NCITS meetings (plus their organisation must be a member in good standing). Then you can attend three more international meeting a year, doing even more boring work, but often in interesting destinations, though as far as I could tell, all of the places I went looked like the inside of a conference room.
So it is a commitment of time and money, but anyone who wants to put in the effort can have a say on what standards are formed.
The people who signed the AUP and then spammed owe you $500. The credit card companies won't collect this debt for you (even though it seems they should). So take the spammers to small claims court - you have a signed agreement and proof they violated it - just because the CC companies won't enforce it doesn't mean it's not a real debt, and small claims is designed for collecting on debts like this. No lawyers needed.
I don't have depth perception: my eyes point in different directions and the images aren't fused. This means that any "3D" display technology that depends on stereo fusion just doesn't work. About 10% of the population has the same sort of problem as me. The Actuality display is real 3D, not cheating like stereo - you can move your head around to see the backs of things.
With the patient inside an MRI machine, they'd have to use unconventional surgical instruments: anything metallic would be really hard to control in the magnetic field. Scalpels made of obsidian?
Playing this kind of odds game is definitely risky. Once the money is invested, treat it as gone and then you'll be pleasantly surprised if you get anything at all back.
Rapid growth happens in small companies, not large ones, and that's where the greatest returns are to be found.
On the other hand, it does have some access controls built in, so you can control what IP addresses (and ports, I think) different internal IP addresses have access to. It might be possible to configure this in such a way that IP addresses handed out via DHCP are prohibited from talking to anything else on the internal network. This would be a stretch since those access controls were designed for things like "don't let Junior's computer connect to FTP", so might not be up to the task of protecting different machines "inside" the firewall from each other.
Making this work is quite a feat, but don't confuse it with good OS design.