Let's take this the next step. I proposed a system a couple of years ago that would eliminate the single point of failure of these systems. I propose that the system *never* write any of the bits of a data element to persistent store. Instead, use a probabalistic Markovian algorithm to immediately re-multicast the data to a number of other hosts in the system, thereby keeping the data always in the *net* (i.e. the routers), and never in the systems themselves. If you want a particular file, your client sends a query in the form of a bias-inducing message that propogates through the system as a diffusing computating. In essence, the bits you are interested in end up in your lap with higher probability, so you just wait a time interval (probably O(lg d*h) where d is # of data elements and h is # hosts participating, based on my back-of-the-envelope calculations given the algorithm I've played with) and the bits happen to show up on your computer and your client reassembles the data appropriately. This is similar to the software distribution multicast work out of MSFT that uses FEC (Forward Error Correction). See GemmelSchoolerGray99 at http://www.cs.caltech.edu/~schooler/papers/FcastTR -99-14.ps for more information. Joseph R. Kiniry
http://www.cs.caltech.edu/~kiniry/
California Institute of Technology
I disagree with several of your assertions, and, as a long-time arcade operator and pinball owner, I should know. Pins are no heavier than the standard vid (still takes two to move either or a lift), they are the same complexity as a vid (you fix EM parts about the same, though the parts you fix in a pin are typically better documented and more "obvious"), and *contrary to what most believe*, a pin *can* be up 100% of the time. The problem is that nearly every arcade owner/manager I've ever known is an *idiot* and doesn't know how to keep the vids running, let alone care about the 15 minutes a week upkeep that a pin requires. My arcades, at their peak (in 1992), had a total of some 50-60 pins all running beautifully. These days, I only maintain a couple of personal machines that I set up at Caltech, and while I love playing my machines, I haven't seen a decently maintained pin at an arcade in over five years.
Final word: the problem isn't the machines or the means, it is the owners and operators.
Joseph R. Kiniry http://www.cs.caltech.edu/~kiniry/ California Institute of Technology
There is a new language available that is specifically designed for introductory instruction in programming. It is called JJ and is a procedural and object-based (you grow through models), simple programming language that focuses on constructs and correctness. It is web-based, so no need to download anything, has lots of documentation, and is geared specifically toward beginners. More information can be found at here
Disclaimer: I was a language designer for JJ. Joseph R. Kiniry http://www.cs.caltech.edu/~kiniry/ California Institute of Technology
I highly suggest (and force on my web admins and designers) the new W3C Recommendation for building tools that support sites that provide accessible content. Such guidelines can be utilized by hand-designers to create aesthetic, useable, and accessible web experiences for all. Every time I access a web page that provides a poor user experience for non-mainstream browsers (I used lynx and w3-mode quite a bit), I email a standard letter to the webmaster that provides some guidelines and references for such. Even if 5% of/.ers did that, the world might start shifting toward a better place.
Joseph R. Kiniry http://www.cs.caltech.edu/~kiniry/ California Institute of Technology
I'd had to say that such a comment is a little off-base, given that Pathfinder used a PowerPC-based architecture. I was lucky enough to take a "special topics in engineering" course here at Caltech with all of the folks that were on the Pathfinder project - a great group of bright, driven folks.
Joseph R. Kiniry http://www.cs.caltech.edu/~kiniry/ California Institute of Technology
How many of us who are out in the dating scene are happy with what we see, both with respect to the relative numbers of women in our field(s) and finding a relationship that is one of peers?
I don't mean that I want to find another over-educated computer wiz - I just want to meet someone who is as driven, creative, and good looking as so many of my male friends, if you get my drift.
I certainly am flattered to be the target of these kinds of events, but I'm not interested in someone who is looking for a wealthy mamma's-boy who is never home and is more knowledgable about his kernel rev than his partner's hopes and dreams.
Some firms actually do care about quality, even at the cost of shipping later, having fewer features, etc. Tradeoffs are a part of the game, unless you have something up your sleeve (e.g. process, technology, theory, etc.).
My new firm, KindSoftware, is planning on offering a similar, but more serious guarantee of quality. I had heard in the past of a firm offering a free VW bug for every bug found, but for a startup, that's probably a price too steep. Of course, the less than adventurous path is to offer extended licenses, extra support, etc. for customers who find bugs. I'm of the opinion that something in between is more appropriate.
Any suggestions? Trips to the Bahamas? $1000 cash? A dual processor over-clocked Celeron?
We at Caltech can certainly appreciate a good hack, even from our rival university.
Interesting enough, coincidentally (?), Ditch Day here at Caltech was yesterday as well. While there were many interesting things found all over campus (and plenty of students dressed up like Jedis), nothing compared to this nice hack at MIT.
Slashdot Mirror
Let's take this the next step. I proposed a system a couple of years ago that would eliminate the single point of failure of these systems. I propose that the system *never* write any of the bits of a data element to persistent store. Instead, use a probabalistic Markovian algorithm to immediately re-multicast the data to a number of other hosts in the system, thereby keeping the data always in the *net* (i.e. the routers), and never in the systems themselves. If you want a particular file, your client sends a query in the form of a bias-inducing message that propogates through the system as a diffusing computating. In essence, the bits you are interested in end up in your lap with higher probability, so you just wait a time interval (probably O(lg d*h) where d is # of data elements and h is # hosts participating, based on my back-of-the-envelope calculations given the algorithm I've played with) and the bits happen to show up on your computer and your client reassembles the data appropriately. This is similar to the software distribution multicast work out of MSFT that uses FEC (Forward Error Correction). See GemmelSchoolerGray99 at http://www.cs.caltech.edu/~schooler/papers/FcastTR -99-14.ps for more information.
Joseph R. Kiniry
http://www.cs.caltech.edu/~kiniry/
California Institute of Technology
I disagree with several of your assertions, and, as a long-time arcade operator and pinball owner, I should know. Pins are no heavier than the standard vid (still takes two to move either or a lift), they are the same complexity as a vid (you fix EM parts about the same, though the parts you fix in a pin are typically better documented and more "obvious"), and *contrary to what most believe*, a pin *can* be up 100% of the time. The problem is that nearly every arcade owner/manager I've ever known is an *idiot* and doesn't know how to keep the vids running, let alone care about the 15 minutes a week upkeep that a pin requires. My arcades, at their peak (in 1992), had a total of some 50-60 pins all running beautifully. These days, I only maintain a couple of personal machines that I set up at Caltech, and while I love playing my machines, I haven't seen a decently maintained pin at an arcade in over five years.
Final word: the problem isn't the machines or the means, it is the owners and operators.
Joseph R. Kiniry
http://www.cs.caltech.edu/~kiniry/
California Institute of Technology
Disclaimer: I was a language designer for JJ.
Joseph R. Kiniry
http://www.cs.caltech.edu/~kiniry/
California Institute of Technology
I highly suggest (and force on my web admins and designers) the new W3C Recommendation for building tools that support sites that provide accessible content. Such guidelines can be utilized by hand-designers to create aesthetic, useable, and accessible web experiences for all. Every time I access a web page that provides a poor user experience for non-mainstream browsers (I used lynx and w3-mode quite a bit), I email a standard letter to the webmaster that provides some guidelines and references for such. Even if 5% of /.ers did that, the world might start shifting toward a better place.
Joseph R. Kiniry
http://www.cs.caltech.edu/~kiniry/
California Institute of Technology
I'd had to say that such a comment is a little off-base, given that Pathfinder used a PowerPC-based architecture. I was lucky enough to take a "special topics in engineering" course here at Caltech with all of the folks that were on the Pathfinder project - a great group of bright, driven folks.
Joseph R. Kiniry
http://www.cs.caltech.edu/~kiniry/
California Institute of Technology
I don't mean that I want to find another over-educated computer wiz - I just want to meet someone who is as driven, creative, and good looking as so many of my male friends, if you get my drift.
I certainly am flattered to be the target of these kinds of events, but I'm not interested in someone who is looking for a wealthy mamma's-boy who is never home and is more knowledgable about his kernel rev than his partner's hopes and dreams.
My new firm, KindSoftware, is planning on offering a similar, but more serious guarantee of quality. I had heard in the past of a firm offering a free VW bug for every bug found, but for a startup, that's probably a price too steep. Of course, the less than adventurous path is to offer extended licenses, extra support, etc. for customers who find bugs. I'm of the opinion that something in between is more appropriate.
Any suggestions? Trips to the Bahamas? $1000 cash? A dual processor over-clocked Celeron?
We at Caltech can certainly appreciate a good hack, even from our rival university.
Interesting enough, coincidentally (?), Ditch Day here at Caltech was yesterday as well. While there were many interesting things found all over campus (and plenty of students dressed up like Jedis), nothing compared to this nice hack at MIT.
Best,
Joe Kiniry (Caltech PhD CS '99)
I would suggest one of the Wiki clones or Zope.
I'm personally biased toward Jiki (http://www.jiki.org).