Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
Dateline March 31, 2006

By fsck! in Technology
Sat Apr 07, 2001 at 01:19:14 AM EST
Tags: Hardware (all tags)
Hardware

Hey Dave,

My new computer just arrived! I decided to get the 150k-node FPGA[1] instead of the 100k-node model. It'll be obsolete in 18 months anyway, so why not? Here's the cool part: 8QBit Quantum Coprocessor[2]. Now I can finally get on the PDIS (Pretty Damn Impenetrable Security) Bandwagon for email, like I should have with GPG back in 2001.


For storage, I went with the 768GB Magnetic Ram[3] clip. I've never really needed a lot of storage. I laugh at those gamers with their overclocked 2TB storage clips. Who needs that much space? I'm sure i'll never use more than 512GB anyway.

They were having a deal on displays, three 24" OLED Panels[4] for the price of two. Damn those things are bright. I'm keeping the mouse around for the sentimental factor, but I'll be doing most of my pointing and grunting with the eye-tracking gear I picked up.

Back in 2001, I realised that the plethora of terminal windows on my X11 screen were really conversations between identities and resources. Just like we control conversations with body language and eye contact, I control my telnet[5] sessions with a little CCD[6] camera that tracks what I'm looking at on screen. It only makes sense that the focus on the screen should jive with the focus on my cornea.

I'm so glad that we don't have to worry about wires anymore. Granted I still pop open the case just to stare at the innards, try to hear the hum. Remember back before Bluetooth 3.2, when we had to connect our peripherals with wires. I don't miss the nest of cables behind my desk at all. BTW, I think i have a pretty good idea about how the wireless security (espicially for PANs) works now. i'll write about that later.

--
Jacob Kuntz

[1] Field Programmable Gate Array, Big Brother to the Connection Machine. These will be the processor of the future. running thousands of threads simultaniously with a minimum of heat or power waste.

[2] Why a coprocessor? Quantum operations are slow for 90% of what computers do. That's why you'll never see quantum graphics accelerator. But what they are good at, they are VERY good at. Remember the 80387?

[3] Magnetic Ram uses the direction of spin of electrons to store data, and the data doesn't go away when the power dies. It will be a lot faster than even the best ram chips of today, and a lot smaller. Home computers with multiple TB of data are a not that far off. Nor are they that useful (probably).

[4] Organic Light Emitting Diodes. Brighter, Crisper, Cheaper displays. I'm not an expert here.

[5] Telnet? Yes, Telnet. Hopefully, by 2006 we will have IP6 on 95% of the internet, negating the need for encryption at the transport layer. SSH, HTTPS, TSIG, and TLS are not the cures to our security woes, but good bandaids. The application layer is another story, as servers will still be comprimised, laptops will still be stolen. Well, PDAs. Laptops should be extinct in a few years. Regardless, (PGP|GPG|PDIS) will be around for a long time.

[6] OK, so there will probably be something better than CCD for this task. I wanted to make a point about how small and inexpensive this device would be.

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Poll
Most Wrong Prediction
o FPGA under every desk 8%
o Quantum Coprocessors 42%
o Multi-TB Magnetic Ram Clips 6%
o OLEDs 0%
o IP6 Adoption 8%
o Eye Tracking for Input 21%
o Wireless Peripherals 3%
o Strong Encryption Surviving at All 9%

Votes: 133
Results | Other Polls

Related Links
o Also by fsck!


Display: Sort:
Dateline March 31, 2006 | 38 comments (28 topical, 10 editorial, 0 hidden)
Overestimating growth... (4.16 / 6) (#5)
by Speare on Sat Mar 31, 2001 at 01:52:39 PM EST

It's easy to overestimate the speed of adoption of new technologies. While many of the concepts you mention *are* being developed, they are not likely to be bargain sales at Fry's even in five years.

FPGAs use MORE power, not LESS power, than traditional CPUs. They don't waste as much if used right, but you're gonna still need to have fans and cooling apparatus. The issue is this: a semiconductor transistor or gate generates heat when it is used. A CPU uses only a low percentage of the transistors at any time, making it inefficient use of the space, but a whole desktop machine is under 250 watts. An FPGA reorganizes its logic on the fly, so if efficiently programmed, it can potentially use most of the transistors all at once. Faster execution, much more heat. Go back and read the materials: 'as low power as a hair dryer' implies 500 to 1500 watts. No cooling? Uh huh. And don't get me started on how to efficiently program these things. You're gonna need a CPU to organize the programmable logic process, just like a Cray needs a nursemaid CPU to feed the problems into the system.

Let's see. We had magnetic core memory before. Solid state transistor logic was denser, which is a plus, it didn't have that annoying issue of errors introduced by nearby current crosstalk. Hard drives are magnetic surface memory. They take time to move the head or platters, and they are prone to crashing if contaminated or struck. If we move away from transistor memory, it'll be for something denser and less prone to failure... not backwards.

I hope OLEDs arrive. It's taken ten years for TFT LCDs to get to anything moderately "high resolution," and their production is hampered by bad pixels which crop up. Think liquid organics will bring higher yield than solid substrates? Maybe, but I wouldn't bank on it.

Just because IPv6 includes some concept of encryption doesn't mean I'd rely on it solely. Encrypted data on top of an encrypted transport layer... doubly hard to crack. You're right about 'not a cure;' security cannot be layered on with a paintbrush, no matter what layer on the network it's applied. Security must be managed.
[ e d @ h a l l e y . c c ]

double encryption (none / 0) (#6)
by delmoi on Sat Mar 31, 2001 at 02:09:32 PM EST

Just because IPv6 includes some concept of encryption doesn't mean I'd rely on it solely. Encrypted data on top of an encrypted transport layer... doubly hard to crack.

Actualy, it's easier to crack. The PGP documentation warns against trying to encrypt something twice for better security. (maybe that was just 2x encryption with the same key, though)
--
"'argumentation' is not a word, idiot." -- thelizman
[ Parent ]
Yep (none / 0) (#32)
by fluffy grue on Sat Apr 07, 2001 at 02:55:43 PM EST

Most encryption mechanisms expose a lot when they're done repeatedly, probably due to the fact of the key now being effectively squared, meaning that there's twice as many factors, meaning it's twice as easy to factor. Similarly, 3DES (triple DES) does an encrypt, then decrypt, then encrypt, all with separate keys, rather than three separate encrypts, for a similar reason (having to do with symmetry in the algorithm and other stuff I don't understand). Basically, what it boils down to is that modulo arithmetic (which almost all encryption mechanisms have in common) has a few properties which make things rather interesting with repeated operations.

Of course, if the different layers use different encryption mechanisms, the double encryption is/should be actually twice as strong, assuming there's no really weird interactions between the two layers. For example, I'd trust DES+Blowfish to be as secure as DES and Blowfish put together (assuming they're using different keys). I wouldn't trust DES+DES.
--
"Is not a quine" is not a quine.
I have a master's degree in science!

[ Hug Your Trikuare ]
[ Parent ]

Triple DES (none / 0) (#36)
by Ronin SpoilSpot on Mon Apr 09, 2001 at 08:43:48 AM EST

> Similarly, 3DES (triple DES) does an encrypt,
> then decrypt, then encrypt, all with separate
> keys, rather than three separate encrypts, for a
> similar reason (having to do with symmetry in the
> algorithm and other stuff I don't understand).

Actually, security is not the reason for 3DES
implementations decoding instead of encoding
in the middle iteration, it is so that you can use
them to do single DES too by just putting in the
same key thrice.

DES is so symmetric that it doesn't really matter
if you encode or decode. IIRC, decoding is
equivalent to encoding with the one's complement of
the key

/RS
"This space intentionally left blank"
[ Parent ]
MRAM (none / 0) (#7)
by dennis on Sat Mar 31, 2001 at 02:16:11 PM EST

We had magnetic core memory before.

Actually IBM just recently announced a major breakthrough in magnetic RAM, and they expect volume production of 256MB chips in 2004. Type "IBM magnetic RAM" into Google and you'll get a slew of articles.

[ Parent ]

Nursemaid CPU on Cray? (none / 0) (#8)
by pmk on Sat Mar 31, 2001 at 02:46:01 PM EST

Traditional Cray SMP machines with custom vector processors run their own operating systems and stand alone as hosts on a network. So does the massively parallel T3E system. They also use satellite systems, but only for I/O control.

Two exceptions are the old T3D, which needed a host Y-MP or C-90, and the Tera MTA gear, which also needs a host.



[ Parent ]

power (none / 0) (#10)
by alprazolam on Sat Mar 31, 2001 at 06:22:20 PM EST

FPGA's use significantly less power than microprocessors. In the end it all depends how they build em. FPGA's don't 'reorganize their logic on the fly', but they are programmable. Like I said, depends on how they do it.

[ Parent ]
Re: Power (none / 0) (#27)
by dufke on Sat Apr 07, 2001 at 04:51:55 AM EST

FPGA's use significantly less power than microprocessors.

But isn't that because they are much smaller? I have a hard time beliveing a 40M transitor 1.4Ghz clocked FPGA drawing less power than a P4. Well, unless you turn half of it off.

In applications where power is really critical (cell phones etc.) you don't use FPGAs. You make ASICs (Application Specific Integrated Circuit). And a normal CPU is structured more like an ASIC than an FPGA. (Well, it IS an ASIC, strictly speaking.)
__
I am a Lurker. If you are reading this, I surfaced momentarily.
[ Parent ]

power (none / 0) (#37)
by JonesBoy on Tue Apr 10, 2001 at 09:41:13 AM EST

FPGA's use gobs more power than their ASIC counterparts. When an ASIC changes state, you are talking about a single transistor. When a FPGA changes state, you are talking about dozens of large transistors. I think the topic article poster was thinking of this link:
http://www.cnn.com/2001/TECH/ptech/03/30/langley.supercomputer/index.html

No, they don't reorg their logic on the fly, but they can reorg the logic that their logic represents on the fly.


Speeding never killed anyone. Stopping did.
[ Parent ]
Organic LED displays? (none / 0) (#34)
by meldroc on Sun Apr 08, 2001 at 01:58:37 PM EST

I hope OLEDs arrive. It's taken ten years for TFT LCDs to get to anything moderately "high resolution," and their production is hampered by bad pixels which crop up. Think liquid organics will bring higher yield than solid substrates? Maybe, but I wouldn't bank on it.

The process for manufacturing OLED displays promises to be much simpler than the TFT LCD manufacturing process. TFT LCDs require circuits to be etched on a the glass in a clean-room environment, then aligned with polarized glass layers, the liquid crystal, the RGB color filters for the pixels, the backlight, etc. The TFT circuits are very delicate, and have frequent yield problems.

OLED displays require fewer steps. There aren't as many layers, just the OLED's themselves, and their circuitry, which does not have to be transparent. I remember hearing that these displays can be sprayed onto surfaces using a sort of inkjet printing process.

AFAIK, the only thing holding up OLED displays is the lifespan of the blue LEDs. While red & green LEDs can work for years, blue OLEDs currently have a MTBF of 1000 hours, meaning they'll start to fail in a matter of months. However, that hurdle should be cleared in a couple years.

[ Parent ]

Encryption nit (4.00 / 2) (#9)
by dennis on Sat Mar 31, 2001 at 02:47:20 PM EST

Now I can finally get on the PDIS (Pretty Damn Impenetrable Security) Bandwagon

Quantum computers can break current public-key encryption algorithms if they have a really large number of bits, but I've never heard of them being able to do encryption exceptionally well. "Quantum encryption" actually refers to prevention of eavesdropping--you set up a communications channel in such a way that nobody midstream can do an observation without being detected. No scrambling of data is involved. A neat idea, but hard to see how you could send those packets through routers without breaking it.

"...No scrambling of data is involved..." (4.00 / 1) (#12)
by mjs on Sat Mar 31, 2001 at 11:04:06 PM EST

In theory it is both: not only encrypted but you know if anyone is even listening in (even trying to copy the bitstream.) Of course, that's "in theory" :)

[ Parent ]
Last theory I saw (4.00 / 1) (#13)
by Miniluv on Sun Apr 01, 2001 at 12:08:30 AM EST

Actually, the last article I saw stated that there is no currently feasible quantum crypto scheme.

Also remember, detecting an eavesdropped on a quantum crypto link means nobody talks to anybody. How is this a benefit?

"Its like someone opened my mouth and stuck a fistful of herbs in it." - Tamio Kageyama, Iron Chef 'Battle Eggplant'
[ Parent ]

Why anti-eavesdropping works (none / 0) (#18)
by dennis on Mon Apr 02, 2001 at 06:05:04 PM EST

I forget the details, but the way it works is, you set up a connection where the recipient can only read the bits if they haven't been observed yet. If anyone midstream makes an observation, quantum correlations are disrupted and the connection collapses. This way if your message recipient can read the message, you know by the laws of physics that no one has eavesdropped.

[ Parent ]
Quantum Encryption has been demonstrated over kms (none / 0) (#28)
by Paul Crowley on Sat Apr 07, 2001 at 05:50:34 AM EST

British Telecom have demonstrated quantum encryption schemes working over kilometers of fiber. These schemes are impractical for the Internet only because routing is difficult; there's no problem in principle with using such schemes for fixed point-to-point links, except that it's more expensive than standard equipment, and there's not much point since our existing symmetric primitives are pretty damn strong.

I foresee no need to replace GPG in the foreseeable future, at least not for security reasons.
--
Paul Crowley aka ciphergoth. Crypto and sex politics. Diary.
[ Parent ]

Quantum Crypto != Quantum Comp. (5.00 / 1) (#30)
by ggy2 on Sat Apr 07, 2001 at 09:54:30 AM EST

Quantum computing and quantum cryptography are two entirely different things. Quantum compuers will be able to break PKC _very_ easily, because they are incredibly good at factoring large integers, which is what gives PKC its security.

Quantum crypto, on the other hand, polarize photons (computers use atoms). In a nutshell, there are two places to stor data, on the rectilinear or diagonal axis. Data goes in one, noise in the other. Then the reciever randomly picks on of the two to get his data from. The reciever fist tells the sender how he read the photons, and the sender tells him which are correct. They disclose a subset of their data and compare it. If the error rate is about 1/4, an eavesdropper has listened. This is known because only one axis of each photon can be read, so is the eavesdropper reads a photon (and recreates it), one of the two polarizations will be wrong. One half of the time the reciever will read this axis, so 1/2*1/2=1/4. If there is no eavesdropper, the sender and reciever use the bits as a key for further communication (in the same way as a one time pad).

[ Parent ]
Moore's Law (4.60 / 5) (#11)
by ucblockhead on Sat Mar 31, 2001 at 08:32:28 PM EST

Assuming Moore's law (double every 18 months, X10 every five years) holds roughly for memory, CPUs and hard drive size (which it roughly has in the fifteen years I've been computing professionally) a typical (midrange) new machine will have, in 2006, the equivalent of:
  • 2.5 Gigs of RAM
  • A 400 Gig hard drive
  • a 7 Ghz CPU
Bandwidth over the last ten years has increased at only a tenth the rate of Moore's law and removable media (floppies, CDs, DVDs) change too chaotically to really say.

Speaking of 15 years ago, isn't it cool how we now all use "bubble memory" and 3D display devices, just like they said back then?
-----------------------
This is k5. We're all tools - duxup

Actually moore's law is dead (2.00 / 3) (#16)
by Sheepdot on Mon Apr 02, 2001 at 01:07:38 PM EST

Moore's law finally hit its end this year when the processor speed failed to double. Please don't seem to recognize this and I feel it will be a major flaw in reasoning.

Expect the next big thing to be speeding up BUS speed on motherboards. I'm guessing this will be hot for about 2 years, then we'll go back to processor speed.


[ Parent ]
Yeh, but (4.50 / 2) (#20)
by delmoi on Tue Apr 03, 2001 at 09:58:21 AM EST

The year before that CPU speed went nuts, intel was pimpin' 1 ghz chips when their roadmaps from a year ago were saying 600mhz was going to be the top. And the doubling happens every 18-24 months, not each year.
--
"'argumentation' is not a word, idiot." -- thelizman
[ Parent ]
Addendum... (4.00 / 1) (#23)
by ucblockhead on Tue Apr 03, 2001 at 05:04:17 PM EST

I should point out that the last should be "the equivalent of a 7 Ghz CPU". Raw CPU speeds have only gone up by a factor of about 100, but other changes have made the actual effective speed go up by another factor of ten. In other words, a PIII running at 10 Mhz would benchmark roughly ten times faster than a 286 running at 10 Mhz.

Though one often ignored factor is that this speed change isn't a constant, across the board increase. For instance, fifteen years ago, floating point math was extremely slow, and integer math was literally a couple of magitudes faster. It took 100 times as long to do 42.5 x 34.1 as 23 x 61. But floating point math has now improved to the point where it runs almost identically to integer math.


-----------------------
This is k5. We're all tools - duxup
[ Parent ]

Eye tracking (4.33 / 3) (#14)
by Miniluv on Sun Apr 01, 2001 at 12:12:24 AM EST

I'm not so sure eye-tracking input technology would be particularly useful. If you think about the process, sure it'd be more efficient if all you did was stare at the window you wanted focused for working, but how many of us really do that?

I frequently look around the screen while typing away at a document, and even more extremely will look away from the monitor when someone talks to me, or the television says something worth listening to, or any of a thousand other reasons. I suspect the processing power necessary to interpret this eye focus data in a useable way without causing people to use sledgehammers for minor adjustments would be unreasonably high.

"Its like someone opened my mouth and stuck a fistful of herbs in it." - Tamio Kageyama, Iron Chef 'Battle Eggplant'

tests (none / 0) (#26)
by cpt kangarooski on Sat Apr 07, 2001 at 02:49:22 AM EST

additionally, i seem to recall having read something where this was tested. iirc, it turns out that your eyes are actually tracking all the hell over the place - it just averages out a lot, and you don't consciously notice.

--
All my posts including this one are in the public domain. I am a lawyer. I am not your lawyer, and this is not legal advice.
[ Parent ]
Looking what typing (none / 0) (#29)
by WWWWolf on Sat Apr 07, 2001 at 07:33:16 AM EST

if you think about the process, sure it'd be more efficient if all you did was stare at the window you wanted focused for working, but how many of us really do that?

Well, I mostly look at where I'm working on, but not always.

A week ago or two, I typed in a long quote from a book, for example - damn sure I was looking at the book, not the editor window or keyboard. (I was surprised that I typed such a long quote with so few typos at such speed. Eat that, OCR!)

Sometimes, typing to hidden or mostly-hidden window is also desired... Alt+Tab, here we come.

-- Weyfour WWWWolf, a lupine technomancer from the cold north...


[ Parent ]
it could work... (none / 0) (#35)
by atomic on Mon Apr 09, 2001 at 07:04:05 AM EST

if it is smart enough not to focus in man pages. :)


atomic.

"why did they have to call it UNIX? that's kind of... ewww." -- mom.
[ Parent ]
Eye tracking (none / 0) (#38)
by Gndlf on Tue Apr 17, 2001 at 01:14:05 PM EST

I suspect the processing power necessary to interpret this eye focus data in a useable way without causing people to use sledgehammers for minor adjustments would be unreasonably high.

Canon and Nikon have used eye-tracking in their "top of the line" cameras for years. You just look at the the object you want to be in focus, and the autofocus adjusts automatically.

On the other hand, I suspect that I would find a "feature" like that just as annoying as "focus follows mouse". Bletch.



[ Parent ]
Wireless (3.00 / 1) (#21)
by Devil Ducky on Tue Apr 03, 2001 at 02:42:44 PM EST

I'll miss the wires behind my desk. When it gets cold out I strech out and put my bare feet under them. I know there not actually any warmer but it works for me...

Devil Ducky

Immune to the Forces of Duct Tape
Day trading at it's Funnest
Speaking of wireless... (none / 0) (#31)
by cr0sh on Sat Apr 07, 2001 at 01:38:30 PM EST

Are they going to bring out the "new" Tesla long-distance electricity system?

Or are we going to be using fuel-cells?

Or just buttloads of batteries?

Until something occurs in the area of "wireless" power, we will always have wires (personally, I wish there was a feasible way to just put the dataports in the same cable as the electrical wire - I am thinking something along the lines of a fiber optic bundle, to eliminate electrical interference. Sure, we would still have cables, but the number of them would be drastically reduced!)...

Wireless Power (none / 0) (#33)
by matthead on Sat Apr 07, 2001 at 11:19:11 PM EST

<p>Couldn't you use something similar to photoelectric cells to power the wireless devices? Something that will pick up electromagnetic radiation on all kinds of wavelengths, in order to get as much power as possible. Of course, you could still have a good old cable for backup power, but if this became efficient enough, it might be able to power the computer. Given enough surface area, of course - think flat, thin computers. </p>

<p>Maybe the displays would consume more power, and would still have power cables. As long as the CPU unit (box?) was exposed to enough EM radiation, I don't see why you couldn't design it effieciently enough to run on this alone - given the leaps fsck's already taken, this doesn't seem so improbable.</p>
--
- Matt
I'm at (0.3, -2.5). Where are you?
[ Parent ]
Dateline March 31, 2006 | 38 comments (28 topical, 10 editorial, 0 hidden)
Display: Sort:

kuro5hin.org

[XML]
All trademarks and copyrights on this page are owned by their respective companies. The Rest 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!