The pigeon hole argument is a great way to get people to see the folly of this claim: Given all possible x byte long inputs, these should compress down to y bytes (y<x). The kicker is that there are fewer y byte messages than x byte ones, so
EITHER some of the plaintexts must compress to the same shorter bit string (in which case how do you choose which one to decompress to) -- two pigeons in the same hole
OR some of the plaintexts don't compress at all.
So in either case, the press release is drivel. Caveat investor.
There's a perl package for doing all sorts of nifty optimizations of your input text, replacing four and for syllables by the digit 4, and analogously for other word sounds. Suprisingly readable even at high levels of compression.
freshmeat will have a link for you. I'd probably search under pager.
Hrm. I thought lzw was the sliding window approach, where compressesd text was either verbatim text or a (go back n chars and copy c chars) tuple. The window is how far back you search for a good match. I'm sure there are wonderful dynamic programming algorithms to speed that search up. It sounds more like you are constructing a huffman tree with a non-uniform depth.
Or is that what LZW is? in which case what is the name of that sliding window approach?
How much juice does it take to keep your standar PCs memory refreshed? If you turn down the refresh rate to absolute minimum and perhaps stagger the refreshes to keep current constant, how long could an embedded battery keep say 128MB alive?
Ihave no idea... but someone has to have an [informed] opinion.
Nah, It'll be free. Well, most of it. And definately(!) copy protected.
TV is really just a vehicle for advertising; as such the three most important things to a network are ratings, ratings, and ratings (a-la realestate). Currently they just have a statistical sample due to neilsen ratings about who watches what when. And the sampling is poor.
If they knew who wanted to watch which shows (and clicking on a link to watch a show is alot more WANT than just flipping a channel) and when (if they watched me, they might discover that late-20's grad students like to watch powerpuff girls and NOVA right about the midnight hour).
The copy-protection is necessary so that they get the when as well as the what, else I might go online and grab a few shows and not even watch them. Only if they can enforce streaming only do they know exactly my viewing habits.
The real purpose of all this data is to be able to target me with ads. The better targeted the ads, the more the airtime costs.
Of course, this all doesn't apply for things like HBO which have no ads, but even there they get valuable demographic data that they can use to offset the pricing structure with.
The hollywood probly does smoothing; apparently studios (there was a slashcomment about this a few months back, but since we can't search archived discussions, it's effectively gone) do edge enhancement because 1) most TVs smooth the image, so the two cancel out 2) it makes people think the DVDs are sharper (a-la early CD players being unbearably "bright" in their sound to make make people notice that crystal clear "digital" sound). 1) implies that a good card optimising for monitor display would apply smoothing.
As for the idct, that is just a bunch of cosine waves that need to be multiply-accumulated -- exactly the thing a DSP is good at, so it makes perfect sense to put that as special purpose hardware. However, I do have to add a bit of salt to your claim that all others do it in software, as MAC loops are bread-and-butter for all DSPs, and most video cards have DSPs (I'm guessing about that one).
Maybe ATI are the only ones to have it completely in hardware, as opposed to firmware+DSP, but why would they do that?
In what way is your standard VGA/RGB output not of sufficiently high quality for broadcast? It seems to me that anything that looks ok on a monitor should be way overspec for conversion to NTSC and display on consumer grade TVs. (NB: I'm assuming you use some special purpose hardware for VGA->NTSC conversion, and not just using the TV-out option)
Could you expand on the ways that SGIs are better to say an ATI-All In Wonder (just to pick a card at random, feel free to choose another).
Are you referring to things like walking and jumping (think of the crap stopmotion Terminator in the end of T-1 or the lack of weight the jumping Raptors had in Jurassic Park (*))?
I had always assumed that physical models and evolution would be the way to go -- you model your Velociraptor and then assign a couple of centers of gravity. Then you drop it in your physical model and let it learn how to balance, run, and jump overnight. Basically, you evolve a dynamic control center for it (neural net probably, underneath a subsumption architecture).
The main problem is that when you now want to constrain the movement -- for example you want it to turn its head halfway through a jump, you might have to go and retrain it for that scenario.
Ok, that was all speculation: can anyone with experience in the field comment on how far in the future the above scenario is?
(*) Mind you, Crouching Tiger went to great pains to get exactly that effect in its fight scenes...
Guido Van Rossum's brother apparently did some work (or worked for a company that...) on active postscript fonts. These fonts would update their font metrics pseudorandomly after each character, thus retaining their general shape yet never printing the exact same way twice.
#146 bascially beat me to the punch, but I'd like to reiterate the fact that there is a whole slew of problems that are not in NP (and not in P) but are unlikely to be NP-complete.
Factors is one such problem, as are most other encryption schemes, including blowfish and rijndael (sp?).
For all these problems, if NP = P is proven, we're hosed. But it is fine if Factors is proven not to be NP-Complete.
As far as I can tell, 2 is inaccurate and 3 is not specific to RSA.
2) Factors doesn't have to be NP complete. If I have a P time way of solving Factors, that doesn't help me solve 3-SAT without a bidirectional reduction between them (from 3-SAT to Factors should be straight forward, the reverse is thought unlikely to exist). Factors only has not to be in P, which would require NP != P. which brings us to #3
3) ALL key-based encryption is based on the assumption that P!=NP, as ALL such schemes are in NP. Thus, if a solution demonstrating how to solve any one NP-complete problem (any one NP problem will not suffice) in P time is found, all key based encryption has been broken.
However, your first point is well taken.
Ok, now that I've nitpicked on your points, feel free to return the favor.
Man I loved that gag. Douglas adams excells at setting things up in one half of the book and then waiting until you're off guard to pull them out for maximum effect.
Problem with firing stuff from space to ground is that you need a REALLY big projectile to have anything left when it hits the ground. You might have better luck with high altitude balloons and/or self piloting drones firing target aquiring ammo.
remind me, how does the coil gun accellerate a non-magnetic slug? Does the changing magnetic field somehow induce a current (and hence magnetism) in the slug?
My last physics course was a Long Time Ago (in a Galaxy Far...)
It's less than 20 ft to the ceiling in most cafes, and I bet that more laptops have IR than 802.11. Mind you, then I'd have to sit wiht the laptop perched at an awkward angle so that the ceiling-mounted sensor could see it... but still, that strikes me as a nice Q-n-D solution.
I've been meaning to get our local CCS dept to wire up the local cafe, as a promotional stunt to attract good students, but of course, I haven't bothered actually telling anyone about my plans -- I'm still at the thinking loudly stage.
Re:bottleneck is probably the compression
on
MP3 Recorders?
·
· Score: 2
You can do wonders with an IC. Especially repetetive things like mp3 encoding can be sped up trivially in hardware by just pipelining the encoder (or even just replicating the encoding circuits and buffer).
This is exactly the sort of task a DSP or IC kicks a general CPU's ass at.
I was going to suggest that somebody do somehting with that IR port that every laptop seems to have. They do 2Mbs, right? should be enough for casual web surfing.
These sort of things are best not fought openly. Instead, dual boot. Should there be a suprise inspection, accidentally turn off your laptop and reboot into your clean setup.
Someone once pointed out that while insubbordination and incompetence are about equally effective, one is much harder to prove than the other.
Just assume that your network is always infiltrated. Encrypt all your trafic.
After all, ethernet is a shared media protocol, whether that media be air or cable. Cable is just [slightly] harder for the determined attacker to bug. For example, I could splice into physical cable and plant a tap there, or -- and now I'm speculating -- an inductive pickup could probably record traffic without even harming the cable at all.
This is why PGP has the option to mark a message as decryptable to screen only. While you are [always] completely hosed if your recipient is malicious, this sort of annotation will make it harder for recipients downstream to compromise you if they are merely lacking in clue.
The lesson is that active documents are a bad idea, unless they have extremely well thought out security infrastructures. Does anyone have pointers to such infrastructures?
Let me also ask my standard question that I've been asking since I first heard of web-bugs: how come there isn't a standard for sending out self contained html document clusters -- several linked pages, with all the graphics and files they need to be viewed? You could then use the standard "you are about to view and insecure page" when you click on any external links they might have.
The big question (the register's claims aside) is how many crusoes it takes to replace one Intel whopper.
Several things play in transmeta's favor:
1) server code is fixed. This means that the costs due to code morphing don't need to be paid often. This is akin to java's HotSpot technology -- it needs to run a long time to amortize its startup costs. I expect crusoes to work better in servers than in clients for this reason alone.
2) Because they can effectively hibernate when idle, you only pay for electricity for the cycles you use. If budgets are tight, you could envisage a system with under-specced cooling, and temp monitoring -- this would allow you to run at say 80% of maximum constantly, and at 100% for short periods until the temperature rose too high and the processors were forcibly throttled back. This probably fits many usage scenarios very well.
3) If it takes several crusoes to replace one itanium, those crosoes will also be much more fault tolerant, allowing you to continue at a lower capacity while the faulty one is replaced and rebooted.
Ok, I'll grant you that the problem is lack of overlap between I/O and computation.
I drew the conclusion that the compilation was IO bound because otherwise I couldn't account for the massive speedup from 1 process to three.
But those conclusions were drawn in haste, and I'm not really willing to make any definitive claims to their accuracy. However, it is interesting to note that the straight forward transliteration (no optimisation) of your test program took longer than the [supposedly expensive] optimisation. How much of that is IO we can only speculate.
So I'll conclude that you've backed me down from being convinced about the I/O boundedness, but I'm not completely converted to your argument either.
besides, it's friday night, and the bars are calling...
The transmeta chips *are* low power. For any given desired processing capacity, the transmeta solution will require less power, space, and cooling than the comparable intel solution.
Of course, holding power fixed, you get more computing horsepower than intel. However, most companies decide on the cpu power first, and then accomodate that with space, cooling, and electricity infrastructure.
Slashdot Mirror
The pigeon hole argument is a great way to get people to see the folly of this claim: Given all possible x byte long inputs, these should compress down to y bytes (y<x). The kicker is that there are fewer y byte messages than x byte ones, so
EITHER some of the plaintexts must compress to the same shorter bit string (in which case how do you choose which one to decompress to) -- two pigeons in the same hole
OR some of the plaintexts don't compress at all.
So in either case, the press release is drivel. Caveat investor.
There's a perl package for doing all sorts of nifty optimizations of your input text, replacing four and for syllables by the digit 4, and analogously for other word sounds. Suprisingly readable even at high levels of compression.
freshmeat will have a link for you. I'd probably search under pager.
Hrm. I thought lzw was the sliding window approach, where compressesd text was either verbatim text or a (go back n chars and copy c chars) tuple. The window is how far back you search for a good match. I'm sure there are wonderful dynamic programming algorithms to speed that search up. It sounds more like you are constructing a huffman tree with a non-uniform depth.
Or is that what LZW is? in which case what is the name of that sliding window approach?
How much juice does it take to keep your standar PCs memory refreshed? If you turn down the refresh rate to absolute minimum and perhaps stagger the refreshes to keep current constant, how long could an embedded battery keep say 128MB alive?
Ihave no idea... but someone has to have an [informed] opinion.
Nah, It'll be free. Well, most of it. And definately(!) copy protected.
TV is really just a vehicle for advertising; as such the three most important things to a network are ratings, ratings, and ratings (a-la realestate). Currently they just have a statistical sample due to neilsen ratings about who watches what when. And the sampling is poor.
If they knew who wanted to watch which shows (and clicking on a link to watch a show is alot more WANT than just flipping a channel) and when (if they watched me, they might discover that late-20's grad students like to watch powerpuff girls and NOVA right about the midnight hour).
The copy-protection is necessary so that they get the when as well as the what, else I might go online and grab a few shows and not even watch them. Only if they can enforce streaming only do they know exactly my viewing habits.
The real purpose of all this data is to be able to target me with ads. The better targeted the ads, the more the airtime costs.
Of course, this all doesn't apply for things like HBO which have no ads, but even there they get valuable demographic data that they can use to offset the pricing structure with.
The hollywood probly does smoothing; apparently studios (there was a slashcomment about this a few months back, but since we can't search archived discussions, it's effectively gone) do edge enhancement because 1) most TVs smooth the image, so the two cancel out 2) it makes people think the DVDs are sharper (a-la early CD players being unbearably "bright" in their sound to make make people notice that crystal clear "digital" sound). 1) implies that a good card optimising for monitor display would apply smoothing.
As for the idct, that is just a bunch of cosine waves that need to be multiply-accumulated -- exactly the thing a DSP is good at, so it makes perfect sense to put that as special purpose hardware. However, I do have to add a bit of salt to your claim that all others do it in software, as MAC loops are bread-and-butter for all DSPs, and most video cards have DSPs (I'm guessing about that one).
Maybe ATI are the only ones to have it completely in hardware, as opposed to firmware+DSP, but why would they do that?
In what way is your standard VGA/RGB output not of sufficiently high quality for broadcast? It seems to me that anything that looks ok on a monitor should be way overspec for conversion to NTSC and display on consumer grade TVs. (NB: I'm assuming you use some special purpose hardware for VGA->NTSC conversion, and not just using the TV-out option)
Could you expand on the ways that SGIs are better to say an ATI-All In Wonder (just to pick a card at random, feel free to choose another).
Are you referring to things like walking and jumping (think of the crap stopmotion Terminator in the end of T-1 or the lack of weight the jumping Raptors had in Jurassic Park (*))?
I had always assumed that physical models and evolution would be the way to go -- you model your Velociraptor and then assign a couple of centers of gravity. Then you drop it in your physical model and let it learn how to balance, run, and jump overnight. Basically, you evolve a dynamic control center for it (neural net probably, underneath a subsumption architecture).
The main problem is that when you now want to constrain the movement -- for example you want it to turn its head halfway through a jump, you might have to go and retrain it for that scenario.
Ok, that was all speculation: can anyone with experience in the field comment on how far in the future the above scenario is?
(*) Mind you, Crouching Tiger went to great pains to get exactly that effect in its fight scenes...
grin!
...) on active postscript fonts. These fonts would update their font metrics pseudorandomly after each character, thus retaining their general shape yet never printing the exact same way twice.
Guido Van Rossum's brother apparently did some work (or worked for a company that
Kinda neat.
You can probably chase it down from python.org
hrm. 'till the end of the world?
(re: the sig, in case you're wondering)
#146 bascially beat me to the punch, but I'd like to reiterate the fact that there is a whole slew of problems that are not in NP (and not in P) but are unlikely to be NP-complete.
Factors is one such problem, as are most other encryption schemes, including blowfish and rijndael (sp?).
For all these problems, if NP = P is proven, we're hosed. But it is fine if Factors is proven not to be NP-Complete.
As far as I can tell, 2 is inaccurate and 3 is not specific to RSA.
2) Factors doesn't have to be NP complete. If I have a P time way of solving Factors, that doesn't help me solve 3-SAT without a bidirectional reduction between them (from 3-SAT to Factors should be straight forward, the reverse is thought unlikely to exist). Factors only has not to be in P, which would require NP != P. which brings us to #3
3) ALL key-based encryption is based on the assumption that P!=NP, as ALL such schemes are in NP. Thus, if a solution demonstrating how to solve any one NP-complete problem (any one NP problem will not suffice) in P time is found, all key based encryption has been broken.
However, your first point is well taken.
Ok, now that I've nitpicked on your points, feel free to return the favor.
Dirk Gently's Detective Agency.
Man I loved that gag. Douglas adams excells at setting things up in one half of the book and then waiting until you're off guard to pull them out for maximum effect.
Problem with firing stuff from space to ground is that you need a REALLY big projectile to have anything left when it hits the ground. You might have better luck with high altitude balloons and/or self piloting drones firing target aquiring ammo.
remind me, how does the coil gun accellerate a non-magnetic slug? Does the changing magnetic field somehow induce a current (and hence magnetism) in the slug?
...)
My last physics course was a Long Time Ago (in a Galaxy Far
It's less than 20 ft to the ceiling in most cafes, and I bet that more laptops have IR than 802.11. Mind you, then I'd have to sit wiht the laptop perched at an awkward angle so that the ceiling-mounted sensor could see it... but still, that strikes me as a nice Q-n-D solution.
I've been meaning to get our local CCS dept to wire up the local cafe, as a promotional stunt to attract good students, but of course, I haven't bothered actually telling anyone about my plans -- I'm still at the thinking loudly stage.
You can do wonders with an IC. Especially repetetive things like mp3 encoding can be sped up trivially in hardware by just pipelining the encoder (or even just replicating the encoding circuits and buffer).
This is exactly the sort of task a DSP or IC kicks a general CPU's ass at.
I was going to suggest that somebody do somehting with that IR port that every laptop seems to have. They do 2Mbs, right? should be enough for casual web surfing.
These sort of things are best not fought openly. Instead, dual boot. Should there be a suprise inspection, accidentally turn off your laptop and reboot into your clean setup.
Someone once pointed out that while insubbordination and incompetence are about equally effective, one is much harder to prove than the other.
Just assume that your network is always infiltrated. Encrypt all your trafic.
After all, ethernet is a shared media protocol, whether that media be air or cable. Cable is just [slightly] harder for the determined attacker to bug. For example, I could splice into physical cable and plant a tap there, or -- and now I'm speculating -- an inductive pickup could probably record traffic without even harming the cable at all.
This is why PGP has the option to mark a message as decryptable to screen only. While you are [always] completely hosed if your recipient is malicious, this sort of annotation will make it harder for recipients downstream to compromise you if they are merely lacking in clue.
The lesson is that active documents are a bad idea, unless they have extremely well thought out security infrastructures. Does anyone have pointers to such infrastructures?
Let me also ask my standard question that I've been asking since I first heard of web-bugs: how come there isn't a standard for sending out self contained html document clusters -- several linked pages, with all the graphics and files they need to be viewed? You could then use the standard "you are about to view and insecure page" when you click on any external links they might have.
The big question (the register's claims aside) is how many crusoes it takes to replace one Intel whopper.
Several things play in transmeta's favor:
1) server code is fixed. This means that the costs due to code morphing don't need to be paid often. This is akin to java's HotSpot technology -- it needs to run a long time to amortize its startup costs. I expect crusoes to work better in servers than in clients for this reason alone.
2) Because they can effectively hibernate when idle, you only pay for electricity for the cycles you use. If budgets are tight, you could envisage a system with under-specced cooling, and temp monitoring -- this would allow you to run at say 80% of maximum constantly, and at 100% for short periods until the temperature rose too high and the processors were forcibly throttled back. This probably fits many usage scenarios very well.
3) If it takes several crusoes to replace one itanium, those crosoes will also be much more fault tolerant, allowing you to continue at a lower capacity while the faulty one is replaced and rebooted.
*clap* *clap* *clap* funniest thing I've read all day, but then I'm writing a research paper...
Ok, I'll grant you that the problem is lack of overlap between I/O and computation.
I drew the conclusion that the compilation was IO bound because otherwise I couldn't account for the massive speedup from 1 process to three.
But those conclusions were drawn in haste, and I'm not really willing to make any definitive claims to their accuracy. However, it is interesting to note that the straight forward transliteration (no optimisation) of your test program took longer than the [supposedly expensive] optimisation. How much of that is IO we can only speculate.
So I'll conclude that you've backed me down from being convinced about the I/O boundedness, but I'm not completely converted to your argument either.
besides, it's friday night, and the bars are calling...
The transmeta chips *are* low power. For any given desired processing capacity, the transmeta solution will require less power, space, and cooling than the comparable intel solution.
Of course, holding power fixed, you get more computing horsepower than intel. However, most companies decide on the cpu power first, and then accomodate that with space, cooling, and electricity infrastructure.