Slashdot Mirror

Just to follow up on my previous post, here are some resources:

• Memetic Algorithms' Home Page
• A Simple Heuristically Guided Search for the Timetable Problem
• (extract from the comp.ai.genetic FAQ)
• An honour's thesis on the topic
• Taiwanese site (in English) with links to papers, etc
• paper and paper available on citeseer

No, this is not a problem for the faint of heart.

Just to follow up on my previous post, here are some resources:

• Memetic Algorithms' Home Page
• A Simple Heuristically Guided Search for the Timetable Problem
• (extract from the comp.ai.genetic FAQ)
• An honour's thesis on the topic
• Taiwanese site (in English) with links to papers, etc
• paper and paper available on citeseer

No, this is not a problem for the faint of heart.

Check out this paper by Chen and Williams. In this work done back at Apple in '93 they describe how to create intermediate camera angles from multiple static images.

Of course, the capacity to fly around the scene in real time had to wait until computers got a lot faster.

thad

The best tools to build parsers and manipulate parsed syntax trees are functional languages, such as OCaml (with its streams parsers, camlp4, ocamllex/ocamlyacc, etc.), or SML, Haskell, etc.

Of course, if you were a LISPer, you'd know that although you have lots of well-known tools to build new parsers, such as Meta or Zebu, the best thing to do about a parser is not to write it, but rather to reuse the builtin extensible universal parser, READ, and its extensible universal unparser, WRITE.

If you spend most of your time writing parsers, you're not just using the wrong tools, you're also using the wrong approach.

Just my .2 mg of e-gold worth...

I'm still considering very fast traveling of the light. If there was an impediment between the 1 meter, I'd consider teleportation.

There's a relativly (pun intended) interesting experiment on NEC Institute's web page. It proves that light can go faster than C. Here for NEC Faster than C webpage

It seems that this beam of light traveled about 3C. However, there's conjecture that it traveled slower, but time dilation made it seem 3C. That's unprovable at the time though.

citeseer is a great thing, and this is a great T-wave overview article:
"Recent Advances in Terahertz Imaging", Mittleman et al

But it isn't even close to a troubling limitation.

...rapid improvements in computer price/performance ratios over the past decade call into question the adequacy of the present UNIX password algorithm.

Given 20 machines and the numbers above, it is probably reasonable to do an exhaustive search of passwords of length 7-8 lower-case letters, 7 lower-case letters and numbers, 6 alpha-numeric characters, 5-6 printable characters, or 5 ASCII characters. The moral is keep your passwords 8 characters long or use lots of unusual characters, but in no circumstance use less than 6 characters. Of course, if the crypt/second/dollar ratio increases by another five orders of magnitude in the next decade, only eight-character passwords that utilize the entire ASCII character set will be immune from brute-force cracking!

Seventh Edition password encryption is long past its use by date. Apple need to do better.

I suspect that you have not been to Asia, and don't understand how crowded a busy train station here can be. Twenty million people commute into Tokyo every day, the vast majority of them by rail. Shinjuku station alone has three million people pass through it every day. If you save a quarter second every time a person goes through a gate, that saves you 208 gate-hours every day, which is like adding 20-30 more gates to the station.

Even the switch from paper tickets (which you slip into a slot, and which pop out another slot at the other end of the gate) to contactless cards has made a noticable difference here in Tokyo.

This photo shows a ticket gate at Shinjuku train station, though not at a particularly busy time. The fellow in the blue jacket looks like he's swiping a Suica card, but he isn't. (This picture was taken before the Suica readers were installed.)

cjs

Strive for minimalism. Less is more. If you can delete code and have it still work, do so.

--
"...lines of code has commonly been found to outperform many of the more complex composite measures of software development." - A. Powell, 1996

David Reed's open spectrum site links a paper showing that if you're willing to put up with some latency, having mobile nodes can make your total wireless network capacity scale up linearly with the number of nodes. (pdf ps). It's a simple store-and-forward scheme. Put this in every car, with a big hard disk, and say goodbye to wireless bandwidth congestion.

Only? Them's mighty strong words, bubba!

You might want to check out Common Lisp. See these links before you make a fool of yourself.

You could also qualify the statement to be something like: `` ... only buzzword-compliant language ... ''

p.s. Notice that I've accepted the proposition that c++ is a high-level language. I probably should have pointed out that that's a questionable assertion. Roll-your-own garbage collection is not high-level!

There seems to be a trend here.

Okay, I meant m less than or equals k. (Slash thought that my less than symbol is a html tag bracket) If you are interested you might want to check out This paper - which surprisingly is old (1988).

-Dracken

How rigorous. Usability pundit picks pet criteria and decides that these are the top HCI labs. Those interested in the real state of the field instead of opinion might take a look at the more rigorous listings available:

Top Research Labs by Topic, 1978 and 1997

Where Researchers Want to Work

BusinessWeek's Top 20 US Research Labs

Google Cache of 1999 US News ranking of User Interaction Grad Schools

MIT Technology Review Corporate R&D Scorecard (Requires subscription)

HCI Academic Article Imapct Rankings

I think that few of the people on avant garde of HCI research take Jacob Neilsen very seriously. He is a usability specialist, not a interface researcher.

Also, given finite memory, Rip Van Winkle can't be cracked:
http://citeseer.nj.nec.com/cachin97unconditional.h tml

It's on citeseer as "Paper as an Analytic Resource for the Design of New Technologies".

Pessimal Algorithms and Simplexity Analysis Read it---you'll like it. Find out the best algorithm to use if your boss makes you sort a list in Paris.

A cached copy of the research paper is available in in several formats from citeseer.

A cached copy of the research paper is available in in several formats from citeseer.

It makes you wonder how much of this is based on theoretical linguistics and formal semantics, and how much is based on good old fashioned statistics and optimization.

Also, if you're a member of IEEE Computing, you can see his publication.

Also, if you're a member of IEEE Computing, you can see his publication.

Though the work presented at crypto 2001 may prove that it's not possible to provide program obfuscation in the general case, some other researchers have shown how to do obfuscating in more restrictive, yet powerfull scenarios.
For example, there is a paper that describes a method to do Function Hiding. This allows to compute a function on an untrusted host. A lot of problems can be modeled that way, and though we may never see methods to provide obfuscation in the general case, it does not rule out the possibility of obfuscating special classes of programs.

(as djbs koobera-server seems to be under hard pressure)

Here you will find mirrors of the original file as well as the document in pdf-format etc:

http://citeseer.nj.nec.com/462633.html

Biham, Shamir - Differential Analysis of DES-Like Cryptosystems.

It contains one of my favourite passages in a crypto paper: "Cryptanalysis of GDES... The special case of q=8 and n=16, which is suggested in [16,18] as a faster and more secure alternative to DES is breakable with just six ciphertexts in a fraction of a second on a personal computer." [and that was a personal computer from 1991 :)].

Well i couldn't get to the original site, but i see that NEC's citeseer has it.

I looked into using a Palm for a bike computer, and concluded that I wasn't very interested. I live in the Seattle area, and I ride in the rain. The BikeBrain solution comes with a plastic protector for your Palm, but it isn't really waterproof.

The good thing of course is that a Palm can capture a lot of data. But just capturing wheel spin data to show speed and distance isn't enough to make me buy either of these solutions.

Last autumn I bought myself a Specialized P.Brain computer. I love it; it collects wheel turn data (like the two Palm solutions) and also altitude and heart rate data. A PC interface lets you capture your data and make pretty charts. You can get a graph showing your speed, altitude, and heart rate plotted against either time or distance. Read more about it here.

The PC download software is for Windows; I'm planning to try to get it working under WINE if I can. The data is stored in some opaque binary format, but you can get the data out with Dan Connelly's Perl script (get it here.

The P.Brain isn't the only data-collecting bike computer. There are other brands. I have heard good things about the Polar XTrainer. There are even computer systems that directly measure your power output; you have a wheel built with a power-measuring hub, and the computer keeps track of power. Pro riders (including Lance Armstrong) use these. For example, the Power-Tap.

steveha

This is why you pretty much need one button control and it would be near a shifter. I.e. go from the statistics screen to a map (assuming you have a GPS receiver plugged in) My GPS tracks where I'm going, as I go along, showing me going down roads, etc. Maybe more helpful when off road, however, GPS work badly on a moving platform in canyons or among redwoods (even works pretty bad inside a small house, so you get an idea how weak the signals are.)

Another option, is the P Brain, don't know much about it, but it gets good reviews and is one button.

Edwin Land (yes, the founder of Polaroid) did work in human color perception, where he showed that two colors could used to create an apparent full color image.

The important things are our visual expectations, as well as the relative intensity of parts of the scene. I can remember a demo from Land where two projectors sufficed to give a full color scene. If part of the image was abstracted, it appeared to be black and white ! This implies that a combination of two colors can, under certain circumstances, appear to be the same as a different combination of three colors. I would suspect that this effect would have to be considered in the vision tests described in the original posts.

A Michigan State U. report on the Land work is available, as is a lot of more recent work, such as this paper by Kobus Barnard.

Yes. You are all alike. You all read fucking Chomsky.

If you don't read Chomsky, I guess that means you don't read much at all. Chomsky is one of the ten most cited authors in history:

"Many are the authors who may wonder is anyone is paying attention to what they write. Professor Noam Chomsky, MIT's preeminent linguistics authority, doesn't have that problem. Recent research on citations in three different citation indices show that Professor Chomsky is one of the most cited individuals in works published in the past 20 years. In fact, his 3,874 citations in the Arts and Humanities Citation Index between 1980 and 1992 make him the most cited living person in that period and the eight most cited source overall -- just behind famed psychiatrist Sigmund Freud and just ahead of philosopher Georg Hegel. Indeed, Professor Chomsky is in illustrious company. The top ten cited sources during the period were: Marx, Lenin, Shakespeare, Aristotle, the Bible, Plato, Freud, Chomsky, Hegel and Cicero."

Also, even though he is a linguist and also known for his political commentary, Professor Chomsky is still among the top 1000 cited authors in Computer Science: ...721. N. Chomsky.

Maybe you should quit watching all thoe sitcom reruns and work on expanding your intellectual horizons. I recommend reading , including those authors with views you don't necessarily agree with. Chomsky is undoubtably one of the most brilliant intellects of our time, the father of modern linguistics, and regardless of whether you like his political views, if you have not read his work in linguistics you only cheating yourself out of a whole universe of wonder.

• Anonymity/Privacy
  
• Infrastructure
  
• Vote buying/Should "surfing"
  
• Vulnerability to hackers
  

Anonymity/Privacy:

This is the point that is the most important about
democracy and therefore also the one that should be stressed the most.
The funny part about this is that it has been known how to do this for
the last 5 years following a paper by R. Cramer, R. Gennaro and
B. Schoenmakers: A
Secure and Optimally Efficient Multi-Authority Election Scheme and
probably before that by others. This is done with homomorphic encryption
which can be compared to putting your vote into a box and locking it.
Then you can take all the locked boxes and combine them into 1 big
(without opening them) and then open the big box and get the result, which
will tell nothing about what each person voted (unless they all voted the
same). And there are alot of other schemes that improve this in certain
ways or make voting in other ways. An example that improves the above
(from exponential decryption to polynomial) is by M. Jurik and I. Damgaard:
A Generalisation, a Simplification
and some Applications of Paillier's Probabilistic Public-Key System.

Infrastructure:

This is probably the thing they're most right about. The problem is that
to make something sufficiently secure you need digital signatures and
that means you need a PKI (public key infrastructure). There are different
ways to do this. First you could require everyone that want's to use
the internet voting to register a public key. Secondly you could distribute
key cards. These 2 are both cumbersome and expensive, which is why I
tend to agree with the article on this point. But there are products out
there that can solve this easily eg.
Cryptomathic which has a
product named easy sign (got to products->Trust Products->easy Sign). In
short this sets up a server that has all the secret keys stored in a
secure way and users can the get the server to sign for them by using
a password system. This is not optimal in terms of security but it is
very close and it grants an very cheap PKI.

Vote buying/Should "surfing":

This is a real problem to e-vote, since you can't really see the person
doing this, but it is also one that with just a little bit of thought
can be done pretty easily. To avoid that anyone gives away their chip
card, password etc. you just bind it to the government information, so
that you can always go to a web site and get your health record, IRS
numbers and stuff like that with the same means of authentication. That
means that if you give it away you will not only give your vote away, but
also alot of personal information. To avoid that you have someone stand
behind you and make sure you vote the right thing, you can introduce
regret offices, where you on the election day can come down and get you
vote changed (getting your signature and the election offices signature)
and then this will override votes that was made over the internet.

As a last point I would like to make is that today it is also possible to
buy votes, either with fake IDs (which might not even be need in certain
places) or give the vote a micro camera and make them take a picture
that they have voted what you wanted.

Vulnerability to hackers:

This I'll only briefly discuss since the 2 papers mentioned in the
anynomity/privacy section actually provide threshold decryption,
which means that if a certain fractions of the servers (>50%) is online
and non-hacked then the correct result can be found. If the servers
is running different hardware and software configurations, it would
take a _extremly_ good hacker to accomplish this in 1 day.
As for DOS there isn't much you can do, but you can always distribute
things which means that the attack will have to be larger to accomplish
total breakdown.

These are my views and I don't say that they are flawless, but they do
give a fair amount of security. There is 1 more point I'll like to
stress which was not in the article since this might actually be an
improvement. The 2 articles mentioned above provides what is known as
universal verifiability, which means that it gives any person the
possibility of checking that the result is correct (given enough
computing power). Now that would be something most americans
would have liked in Florida ?

With better encoding and compression of the document index, a high quality web search engine may fit onto a 7Gb drive of a new PC.

The problem with massive screening systems like these the reverend Thomas Bayes (of Bayes's theorem) is not the detection part, i.e. being able to actually detect all the bad guys, but not drowning in false alarms when doing so. And the base-rate fallacy says that there's not a whole lot you can do about it.

I've developed the argument further in an intrusion detection context see for example The base rate-fallacy and it's implications for the difficulty of intrusion detection, and it's directly applicable here. The article has an introductory example, that explains that under certain conditions a 99% accurate medical test, won't work at all. The references lists a few other papers by Matthews that are well worth a read also.

In short, since there are precious few passengers that are actually "terrorists" for any real definition of the world, the system must be several (perhaps 1x10^5 -- 1x10^6) times better at suppressing false alarms, than at detecting actual terrorist, to avoid the situation where "all" alarms (from a practical standpoint) are false alarms, i.e. the fact that you were flagged says nothing about you being a danger or not.

What's worse of course is that people when faced with such systems start to ignore their output sooner rather than later, and then the system becomes completely useless even from a narrow security perspective.

So, no, it won't work. It could have worked against the "casual" threat, its very existence could have served as a deterrent, but there are hardly any spur-of-the-moment suicide bombers, so, no, scrap that to. It can't work, because Bayes says so.

Actually there are some alternative services for that. Some articles, for example, are available in full from authors' sites and/or services like Citeseer. I don't think Yahoo ever offered it as a free service, though.

a specialized database of some 25 million research documents culled from 7,100 publications, including academic periodicals.

Well, we have one free already here

Me: Wow this is a really good article... *scrolls down*
Salon.com: Did you like this article? Read the rest of it after a subscription...
Me: Nooooooooooooooo! *pulls hair out*

Not that I have anything against subscription services, but I admit I got used to alot of stuff being free.

On the other hand, if you want research documents, try out NEC's Research Index. It's really quite good, I met one of the guys who put it together and talked about the theory behind it. Plus, I got a couple papers in the database.

3am wrote:
Just an aside, this is a fantastic search resource for computer science, math, and related fields:
http://researchindex.com/

This is a great site, everyone, have a peek if you've got a moment. A direct link to the search page follows:

http://citeseer.nj.nec.com/cs

I thought he meant the fdiv bug in the first generation Pentiums.. :)

My mistake, I guess. It's still a little early in the morning here. ;)

"In general, the more links there are to a particular page from other pages, the higher it ranks in Google's hierarchy on that search term."
Most people have a vague idea of how google works. This paper goes into some detail.

A copying garbage collector is actually guaranteed to allocate at least twice as much memory from the operating system then is currently occupied by live objects. Otherwise, it couldn't copy them. Now, imagine what that does for locality of memory, and cache efficiency (This is why Appel's paper comparing gc and stack allocation haven't made most people abandoning the stack just yet).

On the other hand, a mark-sweep garbage collector never moves objects, resulting in external fragmentation (also true of malloc/free), which is also bad for locality of memory and cache efficiency.

You can also use a compacting garbage collector, which works mostly like a mark-sweep garbage collector, but occasionally compacts fragmented memory as well. The downside is complexity (and thus also performance of the compacting phase).

In practice, the problem of external fragmentation isn't really much of a problem with a good allocator using the buddy-system. At least, almost any non-contrived application or benchmark should have less than 10% fragmentation using a good allocator. But that still leaves the problem of having related objects occupy nearby memory locations open (which is needed for good cache efficiency).

Copying garbage collection can be used for that purpose. But if you are using the simple two-space algorithm, you'll find that you are doing the exact opposite of that, the two-space algorithm tends to put related objects far away from each other, due to the depth-first search it uses to find live objects.

So we need another more complex copying collector... But that will again slow down the performance of the collector.

One possible performance neck of a mark-sweep collector is that it needs two passes: one to mark live memory, and one to collect garbage. But the copying collector is no better, first you need to traverse live objects, then you need to copy them somewhere else.

While I still haven't touched on the subject of generational and incremental garbage collection, I think you will already see that writing a good garbage collector is far from trivial. There are no simple cookbook approaches for achieving optimal performance. The only reasonable approach is experimentation, testing, and fine-tuning of parameters. Most real-life garbage collectors use a hybrid of different techniques for different types of objects (distinguished by such attributes as age, size, mutability, etc...)

The problem is very closely related to fair play in wireless ad hoc network routing. In wireless ad hoc networks, nodes forward packets for each other; a selfish node could save battery power and still get their packets routed. At least two papers in the published literature make attempts at this problem:

Enforcing Service Availability in Mobile Ad-Hoc WANs uses secure hardware to achieve this result. Obviously, this makes it open to law-enforcement attack, since the issuer of the hardware is a single point of failure. Also, it's a lot easier to get someone to download something than to buy a piece of secure hardware.

Mitigating Routing Misbehavior in Mobile Ad Hoc Networks doesn't try to stop misbehaving nodes; rather, they try to stop using misbehaving nodes for forwarding. (If you think this scheme is not directly applicable, think of the case of requesting a download of a file you just uploaded.)

Since this is an ongoing area of research, it'll be interesting to see what happens; any workable solution for ad hoc network routing fairness will also ensure p2p fairness. It doesn't work the other way around, since the routing mechanism itself is under attack.

It seems that in a p2p system, including digital signatures in shares, in combination with some kind of reputation system, might be a good way to both achieve fairness and eliminate spam. Maybe allowing leechers in times of excess bandwidth would jumpstart the system (a problem for warezers), and using "moderation point" like things to mod people up and down.

[Disclaimer: I only work in a somewhat related area; I haven't actually considered how one might solve either problem]

The problem is very closely related to fair play in wireless ad hoc network routing. In wireless ad hoc networks, nodes forward packets for each other; a selfish node could save battery power and still get their packets routed. At least two papers in the published literature make attempts at this problem:

Enforcing Service Availability in Mobile Ad-Hoc WANs uses secure hardware to achieve this result. Obviously, this makes it open to law-enforcement attack, since the issuer of the hardware is a single point of failure. Also, it's a lot easier to get someone to download something than to buy a piece of secure hardware.

Mitigating Routing Misbehavior in Mobile Ad Hoc Networks doesn't try to stop misbehaving nodes; rather, they try to stop using misbehaving nodes for forwarding. (If you think this scheme is not directly applicable, think of the case of requesting a download of a file you just uploaded.)

Since this is an ongoing area of research, it'll be interesting to see what happens; any workable solution for ad hoc network routing fairness will also ensure p2p fairness. It doesn't work the other way around, since the routing mechanism itself is under attack.

It seems that in a p2p system, including digital signatures in shares, in combination with some kind of reputation system, might be a good way to both achieve fairness and eliminate spam. Maybe allowing leechers in times of excess bandwidth would jumpstart the system (a problem for warezers), and using "moderation point" like things to mod people up and down.

[Disclaimer: I only work in a somewhat related area; I haven't actually considered how one might solve either problem]

I normally don't reply to anonymous cowards, since they aren't very credible... ;} And if you'd follow the links I provided, you'd find plenty of citations and web links to "credible" sources of information.

However, in this case, I'll make an exception.

Check out:
Complexity International (a refereed journal) Santa Fe Institute (assoc. with Los Alamos Nat. Labs) CiteSeer ResearchIndex of Scientific Papers

If I recall correctly, Adrian commented in his paper from GP96 about the potential use of evolution in accommodating temperature ranges and different chip characteristics. On a somewhat related note, a paper of his on fault tolerance is here.

-- dhilvert@ugcs.caltech.edu

I think one of the biggest problems is that people believe that because they named something clearly, it must automatically be clear and logical to others. More generally than just naming steps of an algorithm, this is a problem with naming commands, functions, variables, etc. I think this paper makes a strong case http://citeseer.nj.nec.com/furnas87vocabulary.html for the naming problem (yes, this is even a problem with experts in a particular field). Names are great, they don't always stand on their own. I'd highly suggest that people read the short section on "armchair" naming. I haven't seen a programmer who wasn't tempted to use this at one point or another.

One alternative is Ken Perlin's QuikWriting, which has been discussed on slashdot and covered by Wired.

"Quikwriting is significantly faster and less stressful to use than Graffiti, and lets you write very quickly without ever picking your stylus up off the surface, although it has the disadvantage that you need to learn a special alphabet. For further info, you can preview a Technote in either PDF or PostScript, which was published at the ACM UIST'98 conference."

Another alternative that builds on Perlin's QuikWriting work, is Francois Guimbretiere's and Terry Winograd's FlowMenus, published at UIST'00.

"We present a new kind of marking menu that was developed for use with a pen device on display surfaces such as large, high resolution, wall-mounted displays. It integrates capabilities of previously separate mechanisms such as marking menus and Quikwriting, and facilitates the entry of multiple commands. While using this menu, the pen never has to leave the active surface so that consecutive menu selections, data entry (text and parameters) and direct manipulation tasks can be integrated fluidly."

I'm currently designing and programming a user interface on the Palm for a remote control application. So I've implemented "Finger Pies", which are simply pie menus that you can use with your finger!

To paraphrase Ben Shneiderman: Finger Pies work well for implementing direct manipulation user interfaces on handheld personal touch screen devices, in which the application provides meaningful, engaging, tightly coupled feedback on the screen, in response to your gesture. By integrating immediate gratification over time, the user enjoys the satisfaction of direct engagement in an immersive experience, and achieves the cognitive resonance of continuous gratification. [My apologies to Ben for the tongue in cheek impression.]

Finger Pies are not meant to replace character input systems like Graffiti, but they are extremely useful and reliable for many applications of handheld input devices, because they're easy enough to use with your finger instead of a pen.

Finger pies are good for reliably selecting between two, four or eight options at a time (which can be nested as pop up submenus), and they're much more robust and resistant to noise than gesture recognition.

One problem with gesture recognition in general, is that it doesn't allow for "reselection" or in-flight refinement and error correction. That is, once you've made a mistake in a gesture, there's no way to change or cancel it, so you will often get characters that you don't mean, and you have to stop what you're doing and erase the mistake.

Pie menus allow you to cancel or change the selection at any time before you commit to the selection, so you can easily browse the menus. So pie menus are most appropriate when there aren't too many items, the items don't change dynamically over time, and when you need to minimize the error rate and selection time.

Most gesture recognition systems are not "self revealing" like pie menus, which can pop up a "map" showing the directions. So pie menus are much easier to learn than gesture recognition, and more appropriate for novice users. Best of all, they naturally train users to "mouse ahead" and select without looking, so they have a smooth, gentle learning curve.

Another advantage of pie menus is that they're not patented or restricted, and there are several freely available open source implementations.

-Don

Penny Lane: "This song was written about the roundabout in liverpool where John and Paul grew up. Half of the song is fact, half is fiction, but most of it is nostalgia. John was starting to write about personal places, and Paul really took this one and ran. "I wrote that the barber had photographs of every head he'd had the pleasure of knowing. Actually, he just had photos of different hair styles. But all the people do stop and say hello." say Paul. Also, "finger pie" is actually an old obscenity in Liverpool. The girls would never thnk of saying the word. It was used in the song as a fun joke for the lads back home. Months after, waitresses in Liverpool had to put up with lads asking for "fish and finger pie." There is also a phallic reference to the "fireman who keeps his fire engine clean." Penny Lane has become a Beatles landmark, and like Blue Jay Way, has it's problems with stolen signs, which are now nicely bolted down. Penny Lane was recorded on December 29, 1966 and released as a single with Strawberry Fields.The song also has a promotional video." -http://members.aol.com/Sumacca/songs.html

If you want to see some of the work of the people cited in the article check out LifeStreams, Data Mountain, and The Task Gallery.

The real question is what task is it that the Desktop doesn't support. Is it finding files, managing files, starting applications, managing windows, etc... I guess if the article had talked more about the whole problem of "metaphors" then maybe I would agree that the Desktop as a metaphor is dead. The metaphor is useful insofar as it allows people to carry over skills from one domain to another. There are many aspects of the Desktop that do not have analogs in physical desktops - for example, menus (to use one from WIMP). This doesn't mean that menus aren't useful, just that people have to learn how they work rather than relying on skills they know from the physical desktop.

The argument in the article seems to be about managing files - the problem is that the Desktop is used for much more than that. That aside, I do think that current file systems usually force me to make up folders and filenames when that isn't really what I care about. I think I would often like to think about my files more in the way of LifeStreams.

I find it a bit absurd that the article implies that folders are an abstraction but stacked desktops are not.

I believe one of the most accurate numerical integration methods is the famous LSODE (Livermore Solver for Ordinary Differential Equations), an implementation of which is available at http://www.netlib.org/odepack/.

Another one is VODE, which is based on LSODE. http://citeseer.nj.nec.com/1230.html. I think both are more accurate than Runge-Kutta.

This section describes how to cryptanalyse GDES [Generalised DES]. The special case ... which is suggested [by Schaumuller-Bichl] as a faster and more secure alternative to DES is breakable with just six [!] ciphertexts in a fraction of a second on a personal computer.

This is one of the early papers in ant-based routing. Quite cool stuff, nice to see that its progressed further.