http://www.zeromq.org/
http://www.zeromq.org/bindings:smalltalk http://www.squeaksource.com/ZeroMQ Here's a Ralph Johnson-ism for you: "Floating point is broken arithmetic". Seems his bete noire as an undergraduate and grad student was floating point. He'd get As in all his courses, but stumble on anything related to floating point. Then one day he figured it was just broken and never had a problem again. Ralph Johnson and Sam Adams from IBM, who is doing the parallel OOP talk today, sat at the back of the Dart presentation, laughed at things Eric Clayberg was saying, and generally were like the bad kids in the class. Clayberg would ask what was going on at the back row and just what was so funny. Ralph would say enigmatic things like "I'm really glad Dart has floating point!" I do get the impression they all know each other. Dart has a type system, but you don't need to use it. And you can use it wrong. int notReallyAnInt = 'Aaron' This works just fine. Apparently the type system "is designed not to be sound". Dart takes the idea of 'isolates' from Erlang. They are processes that don't share state, so I guess they are the opposite of threads. There is an override for definitions, so it's possible to have two isolates, two threads with separate definitions of something as fundamental as Symbol. Dart has no reflection, though they are planning to add 'mirrors' to provide it. This seems to be where you copy an object and make reflective requests of it without touching the original. Clayberg said over and over that they want to make Dart familiar, so arrays start at zero. This kind of thing. There is no eval() as in JavaScript. I'd imagine that cuts down on the metaprogramming capability. eval() allows for the program to generate code that it can execute within itself. It's slow and Dart wants to be speedy. You can snapshot with Dart and it can compile to JavaScript. It's hard to know who Dart is for, what its target market is, if there was an origin story, etc. One person asked whether Dart would replace Java on Android. Or is Dart to replace JavaScript for JavaScript programmers? It does seem plausible that Dart is for Java programmers extending their reach into the browser. The presentation was about features and not strategy, so it wasn't clear to me. Earlier in the day I met Sebastian Heidbrink of Heidbrink Consulting. He lives on Vancouver island and tends a banking client in Germany. Half of his coworkers never realized he moved. When he goes from Vancouver island to Vancouver the city he finds people so frenetic as to be distressing. I find that hilarious. When I go to Vancouver I feel as though I've been forcibly sedated. Sebastian said "I've got Seaside running on ZeroMQ". I had no idea what that meant. He's a big believer of the cloud as an application. If something needs doing, don't add it to your code, fire a socket at something that already does the job. ZeroMQ allows to ... I'm not sure what it does other than it's a networking layer and allows many things to run at once. It's interesting and there are links above with a binding on SqueakSource. Johnny T also know as John Thornton made a presentation on Amber, which looks very polished. He had a slide describing how primitive it is to switch from a text editor to a browser when writing HTML, JS, or CSS, and isn't it wonderful that you can use this great in-browser browser to do it all in the same place the way Amber does. I think this is silly. I have to add precisely two keyboard shortcuts to go back and forth between TextMate and Chrome. I like it. I think Johnny T is prone to be a little susceptible to the 'goly! gee!' nature of things, so that he confuses what is cool for what is practical. His presentation on Amber was good. He likes to do things like code node.js applications using Amber. After his talk he talked with Heidbrink who told him about ZeroMQ. His conversion was instantaneous. The two are going to do a lightning talk on Wednesday about using ZeroMQ for Amber. Today Sam Adams gives a talk on parallel OOP, which I think is going to be about the Roar VM. Chris Muller is going to give a talk on Location-Aware Networks, Context and Business-Intelligence. The Smalltalk Directions panel with Ralph Johnson is today. Again, I have no idea what that'll be about. I figure I'll make it to those three. The others that catch my eye are Using Glorp with New Projects That Need to Access Legacy Data by Mark Grinnell and Object-Centric Profiling: Advanced Visualizations to Tame Wild Execution with Alexander Bergel. There was a boat ride at the end of yesterday on the Gulf. That was fun. There are pelicans here with eighteen inch beaks. The sand beaches are like white sugar. The boat trip out into the bay was an hour and a great way to drink and chat. Chris |
Thanks for the updates. These are really fun reads, so keep 'em coming!
Cheers, - Andreas On 3/20/2012 12:27, Chris Cunnington wrote: > http://www.zeromq.org/ > http://www.zeromq.org/bindings:smalltalk > http://www.squeaksource.com/ZeroMQ > > Here's a Ralph Johnson-ism for you: "Floating point is broken > arithmetic". Seems his bete noire as an undergraduate and grad student > was floating point. He'd get As in all his courses, but stumble on > anything related to floating point. Then one day he figured it was just > broken and never had a problem again. > > Ralph Johnson and Sam Adams from IBM, who is doing the parallel OOP talk > today, sat at the back of the Dart presentation, laughed at things Eric > Clayberg was saying, and generally were like the bad kids in the class. > Clayberg would ask what was going on at the back row and just what was > so funny. Ralph would say enigmatic things like "I'm really glad Dart > has floating point!" I do get the impression they all know each other. > > Dart has a type system, but you don't need to use it. And you can use it > wrong. > > int notReallyAnInt = 'Aaron' > > This works just fine. Apparently the type system "is designed not to be > sound". > Dart takes the idea of 'isolates' from Erlang. They are processes that > don't share state, so I guess they are the opposite of threads. There is > an override for definitions, so it's possible to have two isolates, two > threads with separate definitions of something as fundamental as Symbol. > Dart has no reflection, though they are planning to add 'mirrors' to > provide it. This seems to be where you copy an object and make > reflective requests of it without touching the original. > Clayberg said over and over that they want to make Dart familiar, so > arrays start at zero. This kind of thing. > There is no eval() as in JavaScript. I'd imagine that cuts down on the > metaprogramming capability. eval() allows for the program to generate > code that it can execute within itself. It's slow and Dart wants to be > speedy. > You can snapshot with Dart and it can compile to JavaScript. > > It's hard to know who Dart is for, what its target market is, if there > was an origin story, etc. One person asked whether Dart would replace > Java on Android. Or is Dart to replace JavaScript for JavaScript > programmers? It does seem plausible that Dart is for Java programmers > extending their reach into the browser. The presentation was about > features and not strategy, so it wasn't clear to me. > > Earlier in the day I met Sebastian Heidbrink of Heidbrink Consulting. He > lives on Vancouver island and tends a banking client in Germany. Half of > his coworkers never realized he moved. When he goes from Vancouver > island to Vancouver the city he finds people so frenetic as to be > distressing. I find that hilarious. When I go to Vancouver I feel as > though I've been forcibly sedated. > > Sebastian said "I've got Seaside running on ZeroMQ". I had no idea what > that meant. He's a big believer of the cloud as an application. If > something needs doing, don't add it to your code, fire a socket at > something that already does the job. ZeroMQ allows to ... I'm not sure > what it does other than it's a networking layer and allows many things > to run at once. It's interesting and there are links above with a > binding on SqueakSource. > > Johnny T also know as John Thornton made a presentation on Amber, which > looks very polished. He had a slide describing how primitive it is to > switch from a text editor to a browser when writing HTML, JS, or CSS, > and isn't it wonderful that you can use this great in-browser browser to > do it all in the same place the way Amber does. I think this is silly. I > have to add precisely two keyboard shortcuts to go back and forth > between TextMate and Chrome. I like it. I think Johnny T is prone to be > a little susceptible to the 'goly! gee!' nature of things, so that he > confuses what is cool for what is practical. > > His presentation on Amber was good. He likes to do things like code > node.js applications using Amber. After his talk he talked with > Heidbrink who told him about ZeroMQ. His conversion was instantaneous. > The two are going to do a lightning talk on Wednesday about using ZeroMQ > for Amber. > > Today Sam Adams gives a talk on parallel OOP, which I think is going to > be about the Roar VM. Chris Muller is going to give a talk on > Location-Aware Networks, Context and Business-Intelligence. The > Smalltalk Directions panel with Ralph Johnson is today. Again, I have no > idea what that'll be about. I figure I'll make it to those three. The > others that catch my eye are Using Glorp with New Projects That Need to > Access Legacy Data by Mark Grinnell and Object-Centric Profiling: > Advanced Visualizations to Tame Wild Execution with Alexander Bergel. > > There was a boat ride at the end of yesterday on the Gulf. That was fun. > There are pelicans here with eighteen inch beaks. The sand beaches are > like white sugar. The boat trip out into the bay was an hour and a great > way to drink and chat. > > Chris > > |
In reply to this post by Chris Cunnington
On Tue, 20 Mar 2012 07:27:22 -0400
Chris Cunnington <[hidden email]> wrote: > Here's a Ralph Johnson-ism for you: "Floating point is broken > arithmetic". The pithiest version of this is Anton Householder's "It makes me nervous to fly on airplanes since I know they are designed using floating-point arithmetic." Sweet dreams, -- Christopher Oliver <[hidden email]> |
In reply to this post by Chris Cunnington
On Tue, Mar 20, 2012 at 07:27:22AM -0400, Chris Cunnington wrote:
> Here's a Ralph Johnson-ism for you: "Floating point is broken > arithmetic". Seems his bete noire as an undergraduate and grad > student was floating point. He'd get As in all his courses, but > stumble on anything related to floating point. Then one day he > figured it was just broken and never had a problem again. One of the first things I ever did in Smalltalk was write a program to draw the Mandelbrot set, with my own Complex class. It worked, but I was puzzled about why it was so slow. I couldn't increase the maximum number of iterations above about 7 or 8 before it got unusably slow. For those who don't know the Mandelbrot set, it involves squaring complex numbers repeatedly and adding stuff to them to see if they're one of the numbers that eventually goes out to infinity, or if they just loop around in a finite region. I took a look with the debugger, and discovered that Squeak was doing all of my arithmetic exactly, using rational numbers. When you square rational numbers, they get, on average, twice as big. So I was ending up with these fractions with 200 or 300 digits on both the top and bottom, which slowed the arithmetic down a lot. And every increment to maxiter doubled the number of digits. I added a decimal point to my code and suddenly it was lightning-fast. The decimal point triggered a switch from exact rational arithmetic to contagious, approximate, and O(1) floating-point, which was good enough. Fixed point would have worked just as well, but of course it's broken arithmetic too, just in a slightly different way. Sometimes, late answers are wrong answers, and you just have to choose which wrong answers you want. Kragen |
Kragen Javier Sitaker wrote:
> Sometimes, late answers are wrong answers, and you just have to choose > which wrong answers you want. > > Kragen > A very nice lead-in to Sam Adams' talk.... |
On Tue, Mar 20, 2012 at 12:41:17PM -0400, Alan Knight wrote:
> Kragen Javier Sitaker wrote: > >Sometimes, late answers are wrong answers, and you just have to choose > >which wrong answers you want. > A very nice lead-in to Sam Adams' talk.... Too bad I'm not at the conference! I hope someone will post a summary of the points they found most interesting. |
Free forum by Nabble | Edit this page |