Low power clusters are becoming a bigger topic. Orion multisystems is flogging something similar in a single box, and I think Nasa and Los Alamos have put together some small research clusters aimed at low power. I saw some guys from Los Alamos with a small table top cluster using mini-ITX at ClusterWorld last year.
The idea would be to maximize compute power within the restrictions of one 110V AC power circuit (about 20 amps)with no major HVAC infrastructure.
It DID work in the NeXT world, though. The only problems I came across were when processing some external binary data we had written in Moto 68K format.
NeXT solved this problem with "fat binaries". They can package both x86 and PPC binaries into the same Mach binary file. The user just drags the binary into the applications folder and it works. Back in the day I saw some fat binaries with three architectures: x86, moto 68K, and HP-PA. I think there were some four-architecture binaries around, the above plus Sun SPARC.
Building the fat binaries was dead simple--click a few checkboxes in Project Builder, hit the compile button. The only glitches I came across in the process was when reading some external data files we had that were saved in a binary format.
Of course, the suits at the software vendors might not approve of this capability. They might want to limit users to only a single architecture, so they can sell multiple copies.
Day-to-day operations of the Oil for Food program were run out of the UN Secritariat office. The person in charge of monitoring the program, Benon Sevan, was on the take.
the culture of corruption is so rampant at the UN that no one nation, even the US, can overcome it. The other members of the security council were happy to block any serious investigation of the program when it was in place; top officals in France and Russia were receiving millions, too. The inevitable result of "one nation, one vote" when many of the nations are corrupt oligarchies or dictatorships is still more corruption. It's like working in a committee when a solid majority of the members are out to actively subvert the process.
It's a good basic axiom that if you take a quart of ice-cream and a quart of dog faeces and mix 'em together the result will taste more like the latter than the former. That's the problem with the UN. If you make the free nations and the thug states members of the same club, the danger isn't that they'll meet each other half-way but that the free world winds up going three-quarters, seven-eighths of the way.
1. It's _a_ benefit of text XML that it is sorta kinda human readable. The far greater benefit is that it's interoperable and standardized.
2. Yes, you can compress text XML. However, size is only part of the problem to be solved. Parse speed is another. Another thing people miss is databinding. If you have you often need to not only parse the text XML document but also create Java or other language datastructures in binary formats. This can be much faster in a binary XML format, since you can ship numeric data across the wire in IEEE format. The binary format, in a direct translation from text format, is usually somewhat smaller than the text, roughly 1/3 the size or so. This is bigger than a gzipped text XML file, but remember you can also gzip the binary. I've done some experimental implementations of binary XML (it turns out the technique I used is pretty much the same as what everyone else comes up with) and we found that gzipping the binary XML resulted in a somewhat smaller file than directly gzipping the text XML. creating the binary format apparently adds some structure that the gzip algorithm can exploit.
3. But you can gzip binary formats as well (actually, better than) text formats.
4. Again, size is not always the primary objective of binary XML.
5. Servers with high XML traffic are getting hammered. There are some ways around this, such as using optimized, restricted server-side XML parsers that only work with and are optimized for SOAP, but this introduces its own problems. (Are we using your restricted set of XML or my restricted set of XML?)
It's A proposal for doing XML in a binary format. It's certainly not the W3C choice.
As it stands now, the W3C has a binary XML characterization working group. Their charter is _not_ to provide a binary XML format, but rather characterize the problem, ie define the problem to be solved and what the tradeoffs for various solutions are.
Slashdot Mirror
Low power clusters are becoming a bigger topic. Orion multisystems is flogging something similar in a single box, and I think Nasa and Los Alamos have put together some small research clusters aimed at low power. I saw some guys from Los Alamos with a small table top cluster using mini-ITX at ClusterWorld last year. The idea would be to maximize compute power within the restrictions of one 110V AC power circuit (about 20 amps)with no major HVAC infrastructure.
It just worked. It was really amazing.
NeXT solved this problem with "fat binaries". They can package both x86 and PPC binaries into the same Mach binary file. The user just drags the binary into the applications folder and it works. Back in the day I saw some fat binaries with three architectures: x86, moto 68K, and HP-PA. I think there were some four-architecture binaries around, the above plus Sun SPARC. Building the fat binaries was dead simple--click a few checkboxes in Project Builder, hit the compile button. The only glitches I came across in the process was when reading some external data files we had that were saved in a binary format. Of course, the suits at the software vendors might not approve of this capability. They might want to limit users to only a single architecture, so they can sell multiple copies.
the culture of corruption is so rampant at the UN that no one nation, even the US, can overcome it. The other members of the security council were happy to block any serious investigation of the program when it was in place; top officals in France and Russia were receiving millions, too. The inevitable result of "one nation, one vote" when many of the nations are corrupt oligarchies or dictatorships is still more corruption. It's like working in a committee when a solid majority of the members are out to actively subvert the process.
Mark Steyn has it right:
http://www.telegraph.co.uk/opinion/main.jhtml?x
1. It's _a_ benefit of text XML that it is sorta kinda human readable. The far greater benefit is that it's interoperable and standardized. 2. Yes, you can compress text XML. However, size is only part of the problem to be solved. Parse speed is another. Another thing people miss is databinding. If you have you often need to not only parse the text XML document but also create Java or other language datastructures in binary formats. This can be much faster in a binary XML format, since you can ship numeric data across the wire in IEEE format. The binary format, in a direct translation from text format, is usually somewhat smaller than the text, roughly 1/3 the size or so. This is bigger than a gzipped text XML file, but remember you can also gzip the binary. I've done some experimental implementations of binary XML (it turns out the technique I used is pretty much the same as what everyone else comes up with) and we found that gzipping the binary XML resulted in a somewhat smaller file than directly gzipping the text XML. creating the binary format apparently adds some structure that the gzip algorithm can exploit. 3. But you can gzip binary formats as well (actually, better than) text formats. 4. Again, size is not always the primary objective of binary XML. 5. Servers with high XML traffic are getting hammered. There are some ways around this, such as using optimized, restricted server-side XML parsers that only work with and are optimized for SOAP, but this introduces its own problems. (Are we using your restricted set of XML or my restricted set of XML?)
It's A proposal for doing XML in a binary format. It's certainly not the W3C choice. As it stands now, the W3C has a binary XML characterization working group. Their charter is _not_ to provide a binary XML format, but rather characterize the problem, ie define the problem to be solved and what the tradeoffs for various solutions are.