Well, actually, no, protons aren't charge
carriers through many things other than acidic
solutions.
The charge carrier is the electrically charged
thing that carries the current by moving from place
to place. In metals, the carriers are typically
all electrons, while the nuclei stay put. I'm surprised that they managed
to get a patent on proton charge-carrying, though:
IIRC the protons ("H+ ions") *are* the charge carriers
through the solution of many conventional liquid
cells. (In lead-acid batteries, the SO_4^2-
ions are doing the charge carrying...)
Now we know why they have unique IDs. From the article:
One user of the MA IV is Framatome Technologies Inc., of Lynchburg,
Va. The company's inspection of steam generators in nuclear power
plants is expedited by the computer maintaining an inventory of test
equipment as the instruments are brought in and leave the site. A
technician scans equipment with a barcode scanner attached to the MA
IV PC, which is worn underneath a radioactivity containment suit.
Trouble is, deliberately splitting markets like that is ILLEGAL under the Sherman Antitrust Act. It's restraint of trade (price fixing). IANAL and I haven't yet heard a good reason for them to be able to get away with it -- but my guess is that the SAA doesn't apply because the divisions are across nations -- the different "zones" are outside the federal courts' jurisdiction.
Sure would make an interesting court case, though...
Oddly enough, my friend Josh Hadler did something almost exactly like this. He wanted to study electromagnetic standing waves, so he build himself a long metallic waveguide, some neon tubes (not wired for lighting -- just long glass tubes
with neon in 'em) and a CHEAP
MICROWAVE OVEN. He pulled the microwave apart
and plugged the clystron into the waveguide,
with a Nalgene water hose crossing through the
waveguide to absorb most of the energy. Made
beautiful standing waves.
The best part was that the store was having a
sale -- buy a microwave, get a free home microwave
inspection (where a guy would come out and test
your oven for RF leaks). Of course Josh had
the guy come out and test the microwave AFTER
he had taken it apart and used it to make his
standing-wave-generator. The guy was scared
shitless but tested the apparatus anyhow.
My point is that your example is a particularly
cogent one about using a product in a manner
for which it's not intened.
Ever use a car as a nutcracker? You jack up
a drive wheel, put it in fourth and put a brick
on the gas. Then you throw nuts into the gap
between the wheel and the ground. Works VERY WELL
with fresh walnuts.
Ever use a stereo as a degausser? You
short a speaker line through a long spool of
wire, ram a bunch of iron things into the
center of the spool, and use the volume knob
to degauss.
Ever use a Craftsman screwdriver in a way for
which it wasn't intended? Did you break it?
Using products in ways in which they weren't
intended is a big part of the American ideal.
If the Wright Brothers hadn't used bicycle parts
in a way for which they SERIOUSLY weren't intended, it might take a lot longer to get
to that ski vacation today...
The non-obvious thing that you're probably missing
is the shared, multi-port memory. For some classes
of problem you absolutely have to have a single
large bolus of memory for the crunching to work well. Progressive problems like 3-D fluid simulation require that each processor work
in more-or-less parallel on each timestep, but be
able to interact with data that's handled by the
other ones. That's the weak part of distributed
computing systems (which have higher latency and
lower transfer rates than dedicated systems with
shared memory), and it's why I didn't consider
the bare-bones Beowulf solution. But, yah, for
many problems the bigass collection of cheap
workstations really is the best way to go.
For a few moments I seriously considered that my
lab might want this beast -- we do a fair amount of
computing on a Beowulf cluster now, and I'm anticipating more need for cycles in years to come.
But even at the opening price of $35,000, the
machine doesn't compete well with commodity
workstations anymore.
I went over to Penguin Computing and priced their
eight-Pentium-III rackmount server with all the
trimmings. It came to $70,000 for a system that
comes surprisingly close to the Cray in power.
The Penguin Computing system has 8 Pentium III
550 MHz processors (with 1MB L2 cache each). If you were
buying a real system, you'd probably try to use the
much nicer Athlon series, but you'd certainly
get something closer to a gigahertz from either
AMD or Intel. Either
processor has multiple pipelines, so you're likely
to get more than 1 FLOPS per MHz of processor
speed in optimized applications. You could expect
perhaps 0.6 - 1.2 GFLOP per processor with this system, or 1-2 GFLOP with a system that used
higher clock speed (Athlons might deliver higher
pipelining multipliers?). Multiply by eight processors, and you
find you should be able to expect 10-16 GFLOPS out
of the rackmount server. The Penguin Computing machine has a fully cross-linked bus, so RAM bus
contention is probably about equally problematic
between the two machines. The PC machine has
2GB of memory (compared to the Cray Y-MP's 4GB),
but this may be extensible (at ~2K/GB).
The Magnus has dual 75GB
hard drives, easily matching the Cray's disk
space, but it also has a gigabit fiber link
and a kickass graphics card, which the Cray lacks.
So the rackmount Magnus system offers comparable
performance to the Cray, at twice the initial bid
for the Cray (a reasonable estimate for the final price).
Now let's consider additional costs. The
Cray requires that unspecified pieces of paper
be signed to satisfy the U.S. government, adding
unknown bureaucratic cost. The Magnus does not.
Shipping and installing the Cray will require
thousands of dollars. The Magnus requires
about a hundred dollars. The Cray requires special
power and a cooling system. The Magnus dissipates
about a kilowatt and hence can get by with
an extra HVAC vent. It does notrequire special
power. The Magnus, dissipating about 1kW of
power, would require perhaps $60/month for electricity; the Cray, with at least a factor of
10 more power, requires at least $600/month (plus
the added cost of cooling the dinosaur pen).
So the Cray sounds fun to get, but (surprise!) it
just doesn't stack up once the inconvenience factor
is added in.
I suppose I shouldn't be surprised.
Even an I-Opener is in the same ballpark these
days as a VAX 11/785 -- at least for memory and
raw FLOPS.
I collect old science-fiction pulps. The other night, I was reading through ``The Gods of Mars''
again (it's book 2 in the Barsoom series). It's the
1965 pocket edition. Among other things, the
Copyright notice says that ``This book shall not... be lent out...''.
The thing is, there's no way to enforce such a
restriction on a physical book, and indeed later
editions don't have the restriction. There's got to be a story there somewhere, if only one could find the right people to interview.
I was an avid Apple fan for over a decade, but
finally gave up this last round of computer purchases. I bought a 5300 when they came out, and
it was a crock of sh*t. My machine was
(I think) a 5300cs. It would crash and/or freeze
at the drop of a hat -- there must've been some
strangeness about the interrupt architecture.
It was slooow compared to comparable, and cheaper,
laptops of the day. It was big and heavy, and
its physical structure was poorly designed. It
sucked juice like there was no tomorrow. About
a month after the warranty expired, the screen
quit. I played around with one more generation
(the 1400) but what ``they'' say is true - Apple
lags a generation or so behind, at least in the
laptop world. When they do manage to get a kickass machine out (like the then-screamingly-fast
G3 PowerBlimp), it's still saddled with a
non-OS that obviates the extra power.
(What good is 200MFLOPS if the whole thing comes
to a screaming halt whenever you issue a SCSI
SEEK request?)
The quality-control issues in the 5300 were
what made up my mind, though it took a while to
get up the gumption to switch systems. I'm using a Sony VAIO running Linux
now, and I'm pretty darned happy with it. I don't
plan to go back.
The real, overriding problem with the next platform wasn't its hardware -- it was the crappy release schedule. At the time it came out, I was
at Reed College, one of the development and prototype centers. We had access to some of the
very first NeXTs to come off the assembly line,
at release level 0.7 or something. By 0.8 I started playing with it because they kicked the pants off of everything else we had for raw power
(this was the fall of 1988 and winter of 1989), and I was learning about fluid dynamic simulation.
The problem was: the graphical interface on the
top was buggy as anything (at least in 0.8). I never used it. That
caused trouble, because when someone sat down and
logged in at the console the average time to live
was only about three minutes before the kernel panicked. That didn't keep me from running my
fluid dynamic simulations on 'em -- a single NeXT
could almost keep pace with a dedicated VAX 11/785 running my fluid code.
Along about April or so of 1989, we got our first Decstation 3100, which in turn kicked the pants off the NeXT boxen. Then it was all over but the shouting: they'd taken so long to get NeXTStep finished that their hardware had fallen a generation behind.
IIRC, NeXT boxen weren't released commercially until fall 1989 or winter 1990 -- by which time,
DEC was already announcing their 5000 series
workstations and SUN had leapfrogged the NeXTs too.
Foonly effect all the way -- they just hadda
keep tweaking to get the perfect graphics system,
and by the time they were done with perfection it was too late. If they'd only been able to go to
full production in the fall of 1988 instead of
1989, we'd have seen NeXT boxen everywhere and they'd have had the market-share oomph to run with the big dogs. C'est la vie.
Anyone remember Kaypro? Wildly successful
luggable-computer company that died for related
reasons...
Aside from pluto being the last unvisited planet in the solar system, its surface layers should contain a record of conditions in the early solar system as the gas giants were coalescing. I work with a group of planetologists who study solar system origins, and some of it's rubbing off on me (I normally study the other end of the solar system -- the big hot thing in the center) as I begin to understand the points of interest.
The issues surrounding PKE have mostly to do with large budget squeezes within NASA, with the long flight time, and with the radiothermal generator stuff. No need to spout about the reasons why the budget is tight -- though astute people will recall that there's some sort of orbital treehouse that's a leetle bit over budget. It's also hard to justify now spending a bunch of money that can't conceivably pay off until after the next president's term is over -- the flight time to Pluto is >8.5 years. (That sounds like a long time, until you realize it's taken over 12 years just to get the project from NASA HQ outside the Capitol Beltway.) Other folks have pointed out that, due to the unexpected outbreak of world peace, there's comparatively little nuclear weapons development going on -- and hence not much Pu-238 to be had. Further, all the reactionaries who tried to prevent the Cassini launch (did any of them actually bother to calculate the worst-case release scenarios?) are still around, and now they're mad as Hell. The protests and legal action tripled the cost of the RTGs in Cassini, and PKE will have similar problems.
Solar Probe will have trouble with Pu as well, but at least that mission has an alternative. Solar panels, oddly enough, won't work -- they'd get too hot to work around Mercury's orbit, and melt a few days after that. The current plan has a couple of different solar flybys happening -- that requires RTGs, which will last long enough to do the job. But NASA could back off to a single-flyby mission. Then a jettisonable set of solar panels would be used during cruise phase. During the flyby, power would be supplied by a bank of chemical batteries. But then the probe would be dead, dead, dead shortly after the last data from the flyby were downlinked to Earth.
Both of these spacecraft concepts would require incredible miniaturization. Our proposal (I helped write one submitted by Southwest Research Institute) has instruments that are about the size and mass of full beer cans.
I am very impressed with wavelet compression in general. The wavelet decompostion of an image isn't unique -- that gives the heuristics a litle more ``wiggle room'' to choose the most optimized representation of the image.
I just finished writing a proposal to NASA for some instruments on the Solar Probe spacecraft. That's a pretty telemetry-constrained mission. We tested a proprietary wavelet-compression algorithm at 50:1 on 14-bit images (yes, that's about a quarter-bit per pixel) and even at that level it's very hard to tell the difference between compressed and uncompressed images with the naked eye. (The algorithm seems to work by quantizing the sizes of features in the image).
At that level of compression, a 30Hz stream of 6bit-per-channel 640x480 images would only require just over 3Mbps of bandwidth -- and that's without taking any advantage of the relationship between frames. It's easy to believe that another factor of 50 could come out of a combination of more aggressive compression and either diferential encoding or 3-D wavelets. We could end up with full-motion, full-rate video being squirted through 60kbps connections.
I was once astonished by the ineptitude of our various national organizations; now I realize that they're simply composed of people, most of whom are just trying to get their jobs done in a timely way.
Scientists (even top secret burn-your-publications-before-submitting-them nuclear fizzicists) are smart enough to outwit almost any security scheme -- especially one thought up by the kind of "experts" who end up in government jobs rather than academia. Witness Wen Ho Lee, who, while under investigation for possibly sharing secrets with the Chinese government, was able (for whatever reason) to smuggle hundreds of megabytes of ultrasecret bomb simulation data through the security barrier onto insecure tapes.
Further, the motivations for leaking data are so numerous -- from carelessness through the simple convenience/laziness factor, right up to giant moral dilemmas (such as was explored in Durenmatt's The Physicists) that it's impossible to address them all. Quis custodiet ipsos custodes?
On some level, it doesn't really matter if someone leaked classified data. There's probably nothing there that can't be computed with Metropolis, Runge-Kutta, and a gazillion-node Beowulf cluster of next year's K6's. The basic principles are an open book, and compute technology is growing too quickly for the radiation transfer and nuclear reaction simulations to be more than a few-year barrier. In the long run, everyone interested knows everything.
"When people treat items of a highly classified nature just like it's ordinary stuff, something's wrong," Sen. Richard Shelby, R-Ala., chairman of the Senate intelligence committee,told NBC News (in the MSNBC writeup). Arguably, the real problem comes when people treat items of a highly classified nature just like they're going to stay secret forever. These kind of debacles are just a wake-up call for those who would pretend otherwise.
Hmmm. With a range of only 3 miles, it seems that the submarine is revealing its position pretty well (within a 6 mile circle) unless the buoy manages to avoid giving away its position, too.
Also -- aren't hydrophones capable of finding accurate bearings even over huge distances underwater? I imagine it's a good bit easier to spot an audio carrier than it is to detect/decode the data stream.
In all, either (A) there's more to this than meets the eye or (B) the military is hoping that the readers are exceptionally dense. Maybe both.
There were a couple of gems in the direction and cinematography, even in this steaming heap. I very much liked the treatment of language (echoes of Hunt for Red October and the russian-language transition). I also liked the treatment of the different atmospheric composition and everyone's need to have some kind of equipment in the other's environment.
Of course, there are many dozens of inconsistencies in the treatment of technology and of the storyline itself. The book is pretty awful, but doesn't go to quite such lengths of ludicrosity. I believe the movie compressed a couple of years of Johnny Goodboy learning about civilisation, into a single week. I don't recall the humans flying jumpjets (or any kind of aircraft) in the book, to name two of the more glaring sillinesses in the movie.
I work in a scientific shop that uses mainly Losedows/Office for internal communications and a mix of Linux/BSD/Solaris for scientific work. I've tried both Applixware 4.x and 5.0 and Sun's Star Office, and keep vmware and a losedows partition handy for better interoperability with the rest of the organization.
StarOffice is pretty darned good for what it is. The applications within the suite seem to play reasonably nice with one another, and configuring printing and mail preferences was reasonably quick. The main problem is that it doesn't have sufficient compatibility with Weird to do joint document development across platforms. I do a lot of joint development, where several people contribute to the development of a single document. When trading documents between Office and Weird, some of the style information in each direction. Worse, Ofice seems not to be able to handle the annotation and correction features of Weird (IMHO a wonderful feature that almost makes up for the rest of the hassles in Weird).
I really dislike StarOffice's strategy of putting everything in a miniature-desktop window. It works, but the user interface is clunky and completely different than the more sane KDE and Gnome worlds.
StarOffice also stores its documents in non human readable format. This is almost inexcusable in today's world of XML and multigigabyte disks, but doesn't affect the suite per se. It just makes it harder to get under the hood and fix-up your documents in (e.g.) emacs.
The bottom line is that Staroffice is serviceable for most things, and might even work well as an overall office suite in a shop full of unix-heads. It is good enough to interoperate with the entrenched M$ products on a level of sharing file attachments for review, but not quite congruent enough with Weird to allow joint document development.
Applixware was a total wash for me. Lots of interesting ideas, but poor implementation in 5.0. I hope that they do significantly better in 5.2. While the word processor, drawing program, presenter, and mail program all seem to work reasonably well, the interfaces are not clearly thought through (for example, it's very tedious to make a reduced-size superscript in the word processor; superscripts are common enough for footnotes, exponentiation, etc. that this should be easy). Further, there are lots of hidden glitches in the code. For example, the mailer apparently uses the shell as glue for the sendmail program; addresses with <focus brokets> cause mail to vanish silently with a single-line error message to the stdout of the applixware process.
The worst of it is that the conversion routines into and out of Applixware do not work at better than an alpha-test level. M$ Weird documents with pictures or more than a minor level of formatting cause the Applixware import filter to format the document in strange ways or to hang and spin indefinitely.
I have not yet figured out how to keep Applixware's watchdog process from stealing new mail out of my mail spool (I'd rather keep my incoming mail in emacs vmail), so I run Applix as little as possible and hope that nothing arrives while I"m working.
Good ideas in Applixware include their integration with the existing X environment and their use of human-readable file formats.
I haven't tried most of the features that were new in 5.0. I wish they had concentrated more on making all the old stuff work right. As it is, they've got a rather nice looking toy, but I wouldn't want to try doing any work with it. There's potential for future releases, but only if they start really polishing details rather than expanding into new features.
Counterrevolving pairs of flywheels don't work well as a substitute for good gimbaling. The reason is that, while the *external* forces are nulled out by the counterrevolving aspect, the *internal* forces remain every bit as huge. Because there's no force feedback to the outside, it becomes VERY EASY to trash the main bearings: even a very slight, effortless rotation of the main case causes incredible, huge (opposing) torques to develop in the two main bearings -- which then stop working (crash, boom, tinkle).
Tell me about it. Much of our unmanned space program is inextricably linked with internet access. When I worked for the SOHO project at NASA/GSFC, several of the internal computers were cracked. Among them was mine -- a science workstation that could've (at the time) been used as a staging area for a more concerted attack on the command computers themselves (thanks to trusted-host protocols). The attackers used a well-known but unpatched hole in IRIX 6.2 (by default, the line printer account had no password). They were content to fire up an IRC server and brag about how kew1 they were -- we were lucky it was a random heist.
Some of the other instruments' actual command computers were compromised in similar ways at other times. If the attackers had known what they were doing, (I think they, too, were script kiddies) they could've sent commands to the spacecraft, a million miles away.
The problem for that project, as for so many, is lack of clear forethought about security and time pressure once the system was installed. We had a heterogeneous network set up by people from something like 10 different countries, and many workstations (mine included) that were administered by the scientists who used them.
The big shock for me, both in my experience at NASA and at other high-technology, high-risk ventures, is that people remain people even if they work for NASA. Folks who are interested in flying spacecraft have little time to install the latest OS patches or to design secure protocols -- they're too busy shooting from the hip, making huge volumes of hastily written code work right, or cranking out the next research paper.
IMHO, we need *less* connectivity, not more, to our spacecraft and their ground systems!
2 meters in 10^-21 seconds? NOT. *light* takes a full 6.7x10^-9 seconds to go that far, and protons (being matter) can't even go quite that fast.
This guy is overestimating the speed of his "life" by a factor of over 100 billion -- a mistake comparable to (but graver than!) claiming that only one human lives on our planet.
One item I didn't see mentioned in the story: if commercial code is closed-source, how can you spot stolen^H^H^H^H^H^Hnoncompliant code? Much harder with a stripped binary...
Heh. Only the U.S. Army would invent Imperial Stormtrooper outfits (technological, mechanized, evil) and refer to them as JEDI (mystical, organic) forces.
The DDoS problem is a classic "arms race" -- the problem is sorting out the "good stuff" from the chaff sent by DDoS attackers. It reminds me of a dialogue in Hofstadter's (in?)famous Godel, Escher, Bach.
In Hofstadter's dialogue, the audiophile Crab tries to impress the Tortoise with his excellent stereo system -- and the Tortoise keeps giving the Crab records that are carefully crafted to destroy the stereo system by resonating with the works. If only the Crab were willing to accept a little bit of harmonic distortion, his stereo would be unable to reproduce the sounds and destroy itself when playing the malicious records!
The point of the dialogue was to illustrate a necessary weakness in logical systems -- but in this case there is a strict analogy with the Internet: the better the system works at distributing information (playing music), the more susceptible it is to DOS attacks (the malicious recordings).
The Crab's ultimate (failed) solution was to design a custom record player that would laser- scan the disks, perform a harmonic analysis, and rearrange its modules with a robotic arm if it detected a destructive harmonic sequence. Of course, the Tortoise supplied a record that would resonate with the robotic arm itself, thereby breaking the stereo.
In our current (DDOS) case, some attacks are preventable with filtering -- but that filtering lowers the efficiency of the system as a whole, consuming more resources per request and making the system more susceptible to sufficiently clever attacks. Some attacks may even aim straight for the filtering scheme itself, aiming to cause false diversion of real traffic Several folks have commented that clever h4x0rs can spoof arbitrarily realistic web traffic. Ultimately, there's no way to know the intent of any given packet, and sophisticated filters will simply fall to more sophisticated spoofing.
I worked at NASA/GSFC for four years. Admittedly, it's about as far (geographically) as you can get from JPL and still be in the U.S. -- but I can't imagine a full coverup by NASA of any sort of major secret. The climate of the organization is just too open for that. Of course, there *are* plenty of PHB's (as in any other large organization) and it's at least possible that some engineer, somewhere, believed that the thrusters wouldn't work. But as for a coverup at high levels? Naah.
On the other hand, LOTS of outsiders seem to think that NASA coverups are a good explanation for everything from dropped telemetry in solar images to the mysterious sounds they hear (in Wisconsin) at four in the morning. Comes with being a high profile organization, I guess.
FWIW, here's the NASA response -- it's a press release that came out this morning.
(Gee, it'd sure be nice if we could use <pre> in our HTML -- these things come out in ASCII...)
NASA'S RESPONSE TO UPI'S MARCH 21 MARS POLAR LANDER STORY
James Oberg of UPI claims that NASA knew there was a problem with the Mars Polar Lander propulsion system prior to the Dec. 3 landing attempt and "withheld this conclusion from the public." NASA categorically denies this charge. Here's what NASA did and what NASA said:
The Stephenson report, phase 1, was released to the public on November 10, 1999 during a press conference at NASA Headquarters.
The report made 11 different references to technical issues or concerns involving the propulsion system and the Entry, Descent and Landing (EDL) sequence.
This issue was specifically addressed in the press conference and in "MPL Observation No. 5" and other public recommendations of the Stephenson Phase 1 report. It was entitled, "Cold Firing of Thrusters," and dealt in detail with the catalyst bed issue cited by Mr. Oberg of UPI in his March 21 story, "NASA Knew Mars Polar Lander Doomed."
Had UPI researched the public documents released on Nov. 10, which have been available online at the NASA Home Page, the reporter would have been able to conclude that NASA did indeed publicly address propulsion issues, and specifically, the propulsion system's "catalyst bed" temperature concern.
Based on this review, NASA knew about the concerns with the propulsion system, NASA took corrective action, and NASA hid nothing from the public. We made our concerns known in early November.
Several failure scenarios have been reported in the press over the last few weeks, including the lander legs microswitch issue. Outlets such as the Denver Post, Space Daily, and National Public Radio's "All Things Considered" have covered this angle. There is nothing new in the UPI report relating to this specific issue. The lander legs issue is among the failure modes we are studying.
Both the Stephenson and Casani (John Casani, retired JPL flight programs head and also director of mission assurance) teams have conducted intensive reviews relating to Mars Polar Lander, and their teams have surfaced no evidence relating to thruster acceptance testing irregularities as alleged by UPI. In fact, members of the review teams are using words like "bunk," "complete nonsense," and "wacko," to describe their reactions to UPI's charge.
I've recently tried both Applix and StarOffice as office packages in a technical/scientific writing environment. A major concern was drop-in compatibility with Microsloth Word: I work with a distributed team of people who enjoy dragging- and-dropping word documents into their mail.
Between the two, I have to say I preferred the Applix philosophy and user interface -- it felt like a successful cross between the traditional GUI platforms and the EMACS windowing philosophy (windows go anywhere; each window has more or less all the functionality of any other). In contrast, StarOffice sticks everything inside a single large "desktop in a window" that acts completely differently than Gnome/KDE/Twm.
Applixware has excellent user support. Let me repeat: Applixware has excellent user support. I have submitted more than ten bug reports and comments, and their technical staff have diligently followed through on every issue with alacrity, competence, and speed. I'm very impressed with these guys' follow-through. On the other hand, Applixware isn't really ready for prime time yet. There are still lots of minor "gotcha" glitches due to sloppy coding (probably because of the rush to market). For example, their olefilter engine for import/export of Microsoft documents doesn't work too well. I exported a 50-cell excel spreadsheet and tried to re-import it, to find that two columns and al the formatting had disappeared. Many microsoft word documents appear to hang the olefilter entirely. Their mailer uses shell script calls, and the address isn't properly escaped from the shell -- so addresses of the form "Joe Blow <blow@foo.com>" disappear semi-silently (the shell picks up the brokets as redirects and spits an error message back to stderr, where you're not likely to see it).
The detailed interface in the Applixware products is just different enough from both my Mac habits and my emacs habits that it's frustrating. For example, neither ^W nor ^X is a key combination for "cut" -- you have to hit F6 or something. Why do vendors seem always to insist on re-inventing keycodes? On the good side, they offer extensive macro changing capabilities; but OTOH I haven't got time to learn a whole new macro editing language just to get my word processor working right (I'm too busy reading/...). There are many broken links in the documentation tree. On the good side, they've done a pretty nice job of writing the drool-proof pages online. The tutorial and users' guide is friendly, informative, understandable, and thorough. On the downside, the grease-stained pages aren't there yet. They need some sort of reference manual where I can quickly get the information I want without wading through paragraphs of carefully written handholding that is edited for 6th-grade reading level.
StarOffice is more polished and its keycodes match convention better. If you've ever used a microsoft product, you'll feel right at home -- it feels like a clone of the windoze environment. The downside is that it feels like a clone of the windoze environment. There is so much that is broken about the styling of the user interface that I must constantly deal with low-grade annoyance while operating it. But I'm much more productive in SO than in AW, mostly because I don't spend nearly as much time trying to figure out what clever paradigm the programmers tried to work out.
Neither of the packages is interoperable enough with Microsoft Word that I can do joint document development with my benighted brethren at our other sites. That's unfortunate, as it means I have to support Word on at least some platform -- and I'd been hoping to get rid of my macs (never have used Windoze). Looks like I'll have to hang on to 'em for at least a while yet.
The sample images are very impressive. It's terrific to see that yet another space age technology is available for everyone.
The commercial availability of these kinds of imaging changes so many things. A few off the top of my head:
Friends disappeared on a camping trip in the Grand Tetons? Ask Ikonos to photograph the area and find them -- provided they're smart enough to make "obvious" signs in the snow.
Environmental groups have a chance to spot illegal construction in places to which they have no legal access.
Virtual tourism -- the images are still a bit expensive for random snapshots(!), but what if (for example) space images of Tiannanmen Square had been available on the internet? Talk about a slashdotted site...
Virtually everybody can spy on world events. Ever wonder what's really going on in Afghanistan?
This last aspect will give world governments more accountability about geopolitical "hot spots". When Joe Sixpack (or, at least, Joe Wealthy Sixpack or Earth First! or Greenpeace or International Amnesty) can produce images better than the ones that caused the Cuban Missile Crisis, it will become very hard for dishonest governments (such as our own!) to get away with certain kinds of lies. Of course, the illuminati aren't particularly stupid and will undoubtedly try to regulate or outlaw this stuff.
In that light, the ``snapper'' of the BBC article is intriguing -- apparently the U.S. government has already outlawed certain kinds of spaceborne photography of Israel? Sheesh, you'd think people would eventually figure out that you can't put the genie back in the bottle. (You turn your back on congress for one session...)
Slashdot Mirror
The charge carrier is the electrically charged thing that carries the current by moving from place to place. In metals, the carriers are typically all electrons, while the nuclei stay put. I'm surprised that they managed to get a patent on proton charge-carrying, though: IIRC the protons ("H+ ions") *are* the charge carriers through the solution of many conventional liquid cells. (In lead-acid batteries, the SO_4^2- ions are doing the charge carrying...)
Sure would make an interesting court case, though...
The best part was that the store was having a sale -- buy a microwave, get a free home microwave inspection (where a guy would come out and test your oven for RF leaks). Of course Josh had the guy come out and test the microwave AFTER he had taken it apart and used it to make his standing-wave-generator. The guy was scared shitless but tested the apparatus anyhow.
My point is that your example is a particularly cogent one about using a product in a manner for which it's not intened.
Ever use a car as a nutcracker? You jack up a drive wheel, put it in fourth and put a brick on the gas. Then you throw nuts into the gap between the wheel and the ground. Works VERY WELL with fresh walnuts.
Ever use a stereo as a degausser? You short a speaker line through a long spool of wire, ram a bunch of iron things into the center of the spool, and use the volume knob to degauss.
Ever use a Craftsman screwdriver in a way for which it wasn't intended? Did you break it?
Using products in ways in which they weren't intended is a big part of the American ideal. If the Wright Brothers hadn't used bicycle parts in a way for which they SERIOUSLY weren't intended, it might take a lot longer to get to that ski vacation today...
The non-obvious thing that you're probably missing is the shared, multi-port memory. For some classes of problem you absolutely have to have a single large bolus of memory for the crunching to work well. Progressive problems like 3-D fluid simulation require that each processor work in more-or-less parallel on each timestep, but be able to interact with data that's handled by the other ones. That's the weak part of distributed computing systems (which have higher latency and lower transfer rates than dedicated systems with shared memory), and it's why I didn't consider the bare-bones Beowulf solution. But, yah, for many problems the bigass collection of cheap workstations really is the best way to go.
I went over to Penguin Computing and priced their eight-Pentium-III rackmount server with all the trimmings. It came to $70,000 for a system that comes surprisingly close to the Cray in power.
The Penguin Computing system has 8 Pentium III 550 MHz processors (with 1MB L2 cache each). If you were buying a real system, you'd probably try to use the much nicer Athlon series, but you'd certainly get something closer to a gigahertz from either AMD or Intel. Either processor has multiple pipelines, so you're likely to get more than 1 FLOPS per MHz of processor speed in optimized applications. You could expect perhaps 0.6 - 1.2 GFLOP per processor with this system, or 1-2 GFLOP with a system that used higher clock speed (Athlons might deliver higher pipelining multipliers?). Multiply by eight processors, and you find you should be able to expect 10-16 GFLOPS out of the rackmount server. The Penguin Computing machine has a fully cross-linked bus, so RAM bus contention is probably about equally problematic between the two machines. The PC machine has 2GB of memory (compared to the Cray Y-MP's 4GB), but this may be extensible (at ~2K/GB). The Magnus has dual 75GB hard drives, easily matching the Cray's disk space, but it also has a gigabit fiber link and a kickass graphics card, which the Cray lacks.
So the rackmount Magnus system offers comparable performance to the Cray, at twice the initial bid for the Cray (a reasonable estimate for the final price).
Now let's consider additional costs. The Cray requires that unspecified pieces of paper be signed to satisfy the U.S. government, adding unknown bureaucratic cost. The Magnus does not. Shipping and installing the Cray will require thousands of dollars. The Magnus requires about a hundred dollars. The Cray requires special power and a cooling system. The Magnus dissipates about a kilowatt and hence can get by with an extra HVAC vent. It does notrequire special power. The Magnus, dissipating about 1kW of power, would require perhaps $60/month for electricity; the Cray, with at least a factor of 10 more power, requires at least $600/month (plus the added cost of cooling the dinosaur pen).
So the Cray sounds fun to get, but (surprise!) it just doesn't stack up once the inconvenience factor is added in.
I suppose I shouldn't be surprised. Even an I-Opener is in the same ballpark these days as a VAX 11/785 -- at least for memory and raw FLOPS.
I collect old science-fiction pulps. The other night, I was reading through ``The Gods of Mars'' again (it's book 2 in the Barsoom series). It's the 1965 pocket edition. Among other things, the Copyright notice says that ``This book shall not ... be lent out ...''.
The thing is, there's no way to enforce such a restriction on a physical book, and indeed later editions don't have the restriction. There's got to be a story there somewhere, if only one could find the right people to interview.
The quality-control issues in the 5300 were what made up my mind, though it took a while to get up the gumption to switch systems. I'm using a Sony VAIO running Linux now, and I'm pretty darned happy with it. I don't plan to go back.
The problem was: the graphical interface on the top was buggy as anything (at least in 0.8). I never used it. That caused trouble, because when someone sat down and logged in at the console the average time to live was only about three minutes before the kernel panicked. That didn't keep me from running my fluid dynamic simulations on 'em -- a single NeXT could almost keep pace with a dedicated VAX 11/785 running my fluid code.
Along about April or so of 1989, we got our first Decstation 3100, which in turn kicked the pants off the NeXT boxen. Then it was all over but the shouting: they'd taken so long to get NeXTStep finished that their hardware had fallen a generation behind.
IIRC, NeXT boxen weren't released commercially until fall 1989 or winter 1990 -- by which time, DEC was already announcing their 5000 series workstations and SUN had leapfrogged the NeXTs too.
Foonly effect all the way -- they just hadda keep tweaking to get the perfect graphics system, and by the time they were done with perfection it was too late. If they'd only been able to go to full production in the fall of 1988 instead of 1989, we'd have seen NeXT boxen everywhere and they'd have had the market-share oomph to run with the big dogs. C'est la vie.
Anyone remember Kaypro? Wildly successful luggable-computer company that died for related reasons...
The issues surrounding PKE have mostly to do with large budget squeezes within NASA, with the long flight time, and with the radiothermal generator stuff. No need to spout about the reasons why the budget is tight -- though astute people will recall that there's some sort of orbital treehouse that's a leetle bit over budget. It's also hard to justify now spending a bunch of money that can't conceivably pay off until after the next president's term is over -- the flight time to Pluto is >8.5 years. (That sounds like a long time, until you realize it's taken over 12 years just to get the project from NASA HQ outside the Capitol Beltway.) Other folks have pointed out that, due to the unexpected outbreak of world peace, there's comparatively little nuclear weapons development going on -- and hence not much Pu-238 to be had. Further, all the reactionaries who tried to prevent the Cassini launch (did any of them actually bother to calculate the worst-case release scenarios?) are still around, and now they're mad as Hell. The protests and legal action tripled the cost of the RTGs in Cassini, and PKE will have similar problems.
Solar Probe will have trouble with Pu as well, but at least that mission has an alternative. Solar panels, oddly enough, won't work -- they'd get too hot to work around Mercury's orbit, and melt a few days after that. The current plan has a couple of different solar flybys happening -- that requires RTGs, which will last long enough to do the job. But NASA could back off to a single-flyby mission. Then a jettisonable set of solar panels would be used during cruise phase. During the flyby, power would be supplied by a bank of chemical batteries. But then the probe would be dead, dead, dead shortly after the last data from the flyby were downlinked to Earth.
Both of these spacecraft concepts would require incredible miniaturization. Our proposal (I helped write one submitted by Southwest Research Institute) has instruments that are about the size and mass of full beer cans.
I just finished writing a proposal to NASA for some instruments on the Solar Probe spacecraft. That's a pretty telemetry-constrained mission. We tested a proprietary wavelet-compression algorithm at 50:1 on 14-bit images (yes, that's about a quarter-bit per pixel) and even at that level it's very hard to tell the difference between compressed and uncompressed images with the naked eye. (The algorithm seems to work by quantizing the sizes of features in the image).
At that level of compression, a 30Hz stream of 6bit-per-channel 640x480 images would only require just over 3Mbps of bandwidth -- and that's without taking any advantage of the relationship between frames. It's easy to believe that another factor of 50 could come out of a combination of more aggressive compression and either diferential encoding or 3-D wavelets. We could end up with full-motion, full-rate video being squirted through 60kbps connections.
I was once astonished by the ineptitude of our various national organizations; now I realize that they're simply composed of people, most of whom are just trying to get their jobs done in a timely way.
Scientists (even top secret burn-your-publications-before-submitting-them nuclear fizzicists) are smart enough to outwit almost any security scheme -- especially one thought up by the kind of "experts" who end up in government jobs rather than academia. Witness Wen Ho Lee, who, while under investigation for possibly sharing secrets with the Chinese government, was able (for whatever reason) to smuggle hundreds of megabytes of ultrasecret bomb simulation data through the security barrier onto insecure tapes.
Further, the motivations for leaking data are so numerous -- from carelessness through the simple convenience/laziness factor, right up to giant moral dilemmas (such as was explored in Durenmatt's The Physicists ) that it's impossible to address them all. Quis custodiet ipsos custodes?
On some level, it doesn't really matter if someone leaked classified data. There's probably nothing there that can't be computed with Metropolis, Runge-Kutta, and a gazillion-node Beowulf cluster of next year's K6's. The basic principles are an open book, and compute technology is growing too quickly for the radiation transfer and nuclear reaction simulations to be more than a few-year barrier. In the long run, everyone interested knows everything.
"When people treat items of a highly classified nature just like it's ordinary stuff, something's wrong," Sen. Richard Shelby, R-Ala., chairman of the Senate intelligence committee,told NBC News (in the MSNBC writeup). Arguably, the real problem comes when people treat items of a highly classified nature just like they're going to stay secret forever. These kind of debacles are just a wake-up call for those who would pretend otherwise.
Hmmm. With a range of only 3 miles, it seems that the submarine is revealing its position pretty well (within a 6 mile circle) unless the buoy manages to avoid giving away its position, too.
Also -- aren't hydrophones capable of finding accurate bearings even over huge distances underwater? I imagine it's a good bit easier to spot an audio carrier than it is to detect/decode the data stream.
In all, either (A) there's more to this than meets the eye or (B) the military is hoping that the readers are exceptionally dense. Maybe both.
Of course, there are many dozens of inconsistencies in the treatment of technology and of the storyline itself. The book is pretty awful, but doesn't go to quite such lengths of ludicrosity. I believe the movie compressed a couple of years of Johnny Goodboy learning about civilisation, into a single week. I don't recall the humans flying jumpjets (or any kind of aircraft) in the book, to name two of the more glaring sillinesses in the movie.
StarOffice is pretty darned good for what it is. The applications within the suite seem to play reasonably nice with one another, and configuring printing and mail preferences was reasonably quick. The main problem is that it doesn't have sufficient compatibility with Weird to do joint document development across platforms. I do a lot of joint development, where several people contribute to the development of a single document. When trading documents between Office and Weird, some of the style information in each direction. Worse, Ofice seems not to be able to handle the annotation and correction features of Weird (IMHO a wonderful feature that almost makes up for the rest of the hassles in Weird).
I really dislike StarOffice's strategy of putting everything in a miniature-desktop window. It works, but the user interface is clunky and completely different than the more sane KDE and Gnome worlds.
StarOffice also stores its documents in non human readable format. This is almost inexcusable in today's world of XML and multigigabyte disks, but doesn't affect the suite per se. It just makes it harder to get under the hood and fix-up your documents in (e.g.) emacs.
The bottom line is that Staroffice is serviceable for most things, and might even work well as an overall office suite in a shop full of unix-heads. It is good enough to interoperate with the entrenched M$ products on a level of sharing file attachments for review, but not quite congruent enough with Weird to allow joint document development.
Applixware was a total wash for me. Lots of interesting ideas, but poor implementation in 5.0. I hope that they do significantly better in 5.2. While the word processor, drawing program, presenter, and mail program all seem to work reasonably well, the interfaces are not clearly thought through (for example, it's very tedious to make a reduced-size superscript in the word processor; superscripts are common enough for footnotes, exponentiation, etc. that this should be easy). Further, there are lots of hidden glitches in the code. For example, the mailer apparently uses the shell as glue for the sendmail program; addresses with <focus brokets> cause mail to vanish silently with a single-line error message to the stdout of the applixware process.
The worst of it is that the conversion routines into and out of Applixware do not work at better than an alpha-test level. M$ Weird documents with pictures or more than a minor level of formatting cause the Applixware import filter to format the document in strange ways or to hang and spin indefinitely.
I have not yet figured out how to keep Applixware's watchdog process from stealing new mail out of my mail spool (I'd rather keep my incoming mail in emacs vmail), so I run Applix as little as possible and hope that nothing arrives while I"m working.
Good ideas in Applixware include their integration with the existing X environment and their use of human-readable file formats.
I haven't tried most of the features that were new in 5.0. I wish they had concentrated more on making all the old stuff work right. As it is, they've got a rather nice looking toy, but I wouldn't want to try doing any work with it. There's potential for future releases, but only if they start really polishing details rather than expanding into new features.
Counterrevolving pairs of flywheels don't work well as a substitute for good gimbaling. The reason is that, while the *external* forces are nulled out by the counterrevolving aspect, the *internal* forces remain every bit as huge. Because there's no force feedback to the outside, it becomes VERY EASY to trash the main bearings: even a very slight, effortless rotation of the main case causes incredible, huge (opposing) torques to develop in the two main bearings -- which then stop working (crash, boom, tinkle).
Some of the other instruments' actual command computers were compromised in similar ways at other times. If the attackers had known what they were doing, (I think they, too, were script kiddies) they could've sent commands to the spacecraft, a million miles away.
The problem for that project, as for so many, is lack of clear forethought about security and time pressure once the system was installed. We had a heterogeneous network set up by people from something like 10 different countries, and many workstations (mine included) that were administered by the scientists who used them.
The big shock for me, both in my experience at NASA and at other high-technology, high-risk ventures, is that people remain people even if they work for NASA. Folks who are interested in flying spacecraft have little time to install the latest OS patches or to design secure protocols -- they're too busy shooting from the hip, making huge volumes of hastily written code work right, or cranking out the next research paper.
IMHO, we need *less* connectivity, not more, to our spacecraft and their ground systems!
This guy is overestimating the speed of his "life" by a factor of over 100 billion -- a mistake comparable to (but graver than!) claiming that only one human lives on our planet.
One item I didn't see mentioned in the story: if commercial code is closed-source, how can you spot stolen^H^H^H^H^H^Hnoncompliant code? Much harder with a stripped binary...
Heh. Only the U.S. Army would invent Imperial Stormtrooper outfits (technological, mechanized, evil) and refer to them as JEDI (mystical, organic) forces.
The DDoS problem is a classic "arms race" -- the problem is sorting out the "good stuff" from the chaff sent by DDoS attackers. It reminds me of a dialogue in Hofstadter's (in?)famous Godel, Escher, Bach.
In Hofstadter's dialogue, the audiophile Crab tries to impress the Tortoise with his excellent stereo system -- and the Tortoise keeps giving the Crab records that are carefully crafted to destroy the stereo system by resonating with the works. If only the Crab were willing to accept a little bit of harmonic distortion, his stereo would be unable to reproduce the sounds and destroy itself when playing the malicious records!
The point of the dialogue was to illustrate a necessary weakness in logical systems -- but in this case there is a strict analogy with the Internet: the better the system works at distributing information (playing music), the more susceptible it is to DOS attacks (the malicious recordings).
The Crab's ultimate (failed) solution was to design a custom record player that would laser- scan the disks, perform a harmonic analysis, and rearrange its modules with a robotic arm if it detected a destructive harmonic sequence. Of course, the Tortoise supplied a record that would resonate with the robotic arm itself, thereby breaking the stereo.
In our current (DDOS) case, some attacks are preventable with filtering -- but that filtering lowers the efficiency of the system as a whole, consuming more resources per request and making the system more susceptible to sufficiently clever attacks. Some attacks may even aim straight for the filtering scheme itself, aiming to cause false diversion of real traffic Several folks have commented that clever h4x0rs can spoof arbitrarily realistic web traffic. Ultimately, there's no way to know the intent of any given packet, and sophisticated filters will simply fall to more sophisticated spoofing.
I worked at NASA/GSFC for four years. Admittedly, it's about as far (geographically) as you can get from JPL and still be in the U.S. -- but I can't imagine a full coverup by NASA of any sort of major secret. The climate of the organization is just too open for that. Of course, there *are* plenty of PHB's (as in any other large organization) and it's at least possible that some engineer, somewhere, believed that the thrusters wouldn't work. But as for a coverup at high levels? Naah.
On the other hand, LOTS of outsiders seem to think that NASA coverups are a good explanation for everything from dropped telemetry in solar images to the mysterious sounds they hear (in Wisconsin) at four in the morning. Comes with being a high profile organization, I guess.
FWIW, here's the NASA response -- it's a press release that came out this morning.
(Gee, it'd sure be nice if we could use <pre> in our HTML -- these things come out in ASCII...)
NASA'S RESPONSE TO UPI'S MARCH 21 MARS POLAR LANDER STORY
James Oberg of UPI claims that NASA knew there was a problem with the Mars Polar Lander propulsion system prior to the Dec. 3 landing attempt and "withheld this conclusion from the public." NASA categorically denies this charge. Here's what NASA did and what NASA said:
- end -
Between the two, I have to say I preferred the Applix philosophy and user interface -- it felt like a successful cross between the traditional GUI platforms and the EMACS windowing philosophy (windows go anywhere; each window has more or less all the functionality of any other). In contrast, StarOffice sticks everything inside a single large "desktop in a window" that acts completely differently than Gnome/KDE/Twm.
Applixware has excellent user support. Let me repeat: Applixware has excellent user support. I have submitted more than ten bug reports and comments, and their technical staff have diligently followed through on every issue with alacrity, competence, and speed. I'm very impressed with these guys' follow-through. On the other hand, Applixware isn't really ready for prime time yet. There are still lots of minor "gotcha" glitches due to sloppy coding (probably because of the rush to market). For example, their olefilter engine for import/export of Microsoft documents doesn't work too well. I exported a 50-cell excel spreadsheet and tried to re-import it, to find that two columns and al the formatting had disappeared. Many microsoft word documents appear to hang the olefilter entirely. Their mailer uses shell script calls, and the address isn't properly escaped from the shell -- so addresses of the form "Joe Blow <blow@foo.com>" disappear semi-silently (the shell picks up the brokets as redirects and spits an error message back to stderr, where you're not likely to see it).
The detailed interface in the Applixware products is just different enough from both my Mac habits and my emacs habits that it's frustrating. For example, neither ^W nor ^X is a key combination for "cut" -- you have to hit F6 or something. Why do vendors seem always to insist on re-inventing keycodes? On the good side, they offer extensive macro changing capabilities; but OTOH I haven't got time to learn a whole new macro editing language just to get my word processor working right (I'm too busy reading /. ..). There are many broken links in the documentation tree. On the good side, they've done a pretty nice job of writing the drool-proof pages online. The tutorial and users' guide is friendly, informative, understandable, and thorough. On the downside, the grease-stained pages aren't there yet. They need some sort of reference manual where I can quickly get the information I want without wading through paragraphs of carefully written handholding that is edited for 6th-grade reading level.
StarOffice is more polished and its keycodes match convention better. If you've ever used a microsoft product, you'll feel right at home -- it feels like a clone of the windoze environment. The downside is that it feels like a clone of the windoze environment. There is so much that is broken about the styling of the user interface that I must constantly deal with low-grade annoyance while operating it. But I'm much more productive in SO than in AW, mostly because I don't spend nearly as much time trying to figure out what clever paradigm the programmers tried to work out.
Neither of the packages is interoperable enough with Microsoft Word that I can do joint document development with my benighted brethren at our other sites. That's unfortunate, as it means I have to support Word on at least some platform -- and I'd been hoping to get rid of my macs (never have used Windoze). Looks like I'll have to hang on to 'em for at least a while yet.
Actually, it just needs Java and Javascript. I'm running Mozilla , but usually have J&J turned off. Turning 'em back on made the nifty tool work OK.
The sample images are very impressive. It's terrific to see that yet another space age technology is available for everyone.
The commercial availability of these kinds of imaging changes so many things. A few off the top of my head:
This last aspect will give world governments more accountability about geopolitical "hot spots". When Joe Sixpack (or, at least, Joe Wealthy Sixpack or Earth First! or Greenpeace or International Amnesty) can produce images better than the ones that caused the Cuban Missile Crisis, it will become very hard for dishonest governments (such as our own!) to get away with certain kinds of lies. Of course, the illuminati aren't particularly stupid and will undoubtedly try to regulate or outlaw this stuff.
In that light, the ``snapper'' of the BBC article is intriguing -- apparently the U.S. government has already outlawed certain kinds of spaceborne photography of Israel? Sheesh, you'd think people would eventually figure out that you can't put the genie back in the bottle. (You turn your back on congress for one session...)