In Traction - Latest Comments in General

Re: Fun with Python, OpenCV and face detection

daxroc — Wed, 02 Jul 2008 20:09:37 -0000

Nice script , worked fine for me @30fps

Re: Fun with Python, OpenCV and face detection

jozilla — Tue, 01 Jul 2008 05:34:30 -0000

You are right, I noticed yesterday that the cascade file was part of OpenCV. Thanks for clarifying!

Re: Fun with Python, OpenCV and face detection

Nirav Patel — Tue, 01 Jul 2008 01:57:52 -0000

Just to clarify, the Haar cascade file is one of the sample ones that comes with OpenCV. The script I wrote was based on the OpenCV face detection sample too, but using gstreamer instead of OpenCV's HighGUI to interface the webcam (HighGUI didn't support the camera in the XO).

Its good that you got it working though. There is some pretty amazing stuff in OpenCV.

Re: Small update on face detection post

jozilla — Sun, 29 Jun 2008 09:29:55 -0000

Thanks Vahid!

The idea is not new though. Eric Horvitz from Microsoft Research has done impressive research in this area (be sure to check out the movies at the bottom of the page).

Re: Small update on face detection post

vahid — Sun, 29 Jun 2008 07:04:14 -0000

Nice work Jo.
Mixing the nofity and opencv is a great idea in HCI :)

Re: Fun with Python, OpenCV and face detection

jozilla — Sat, 28 Jun 2008 14:53:20 -0000

You're welcome! I'm glad you find it useful.

Re: Fun with Python, OpenCV and face detection

Si — Sat, 28 Jun 2008 12:09:03 -0000

excellent stuff. Many thanks

Re: Demo video of a Smalltalk environment

jozilla — Sun, 25 May 2008 14:27:37 -0000

Hi Ruben, thanks for your feedback!

I already mentioned in my previous blog post that I believe these ideas should be included in our OOP course. I am not involved with the course though. A good time to include this might be when the course material will be revised :-)

Unfortunately I won't make it to the lecture by Professor Daubechies either. I leave for AVI 2008 in Napels on Tuesday together with Jan Meskens to present our paper. So indeed, see you at another CS event :-)

Enjoy your vacation!

Re: Demo video of a Smalltalk environment

Ruben — Sun, 25 May 2008 12:48:56 -0000

What a useful video! I didn't know anything about Smalltalk and found this a perfect introduction. Maybe something to show in courses at UHasselt ?

Anyway i'm looking forward to see some new cool blogposts :-)

Cheers
Ruben

PS : Unfortunately I won't be present at thursday's reading at Uhasselt (Mont Ventoux by bike, here we come :-)), so see you next time!

Re: Back to the future: Smalltalk

Mark Miller — Fri, 23 May 2008 04:24:18 -0000

I'm a little familiar with what you're talking about. From the screenshots I've seen, Dolphin integrated better with Windows in the visual sense. The browser window was just another window on the Windows desktop, for example, as was your app., and any debugger windows. ObjectStudio is the same way. Squeak has its own UI separate from the native UI.

I tried my hand with the .Net Bridge, and it works well if you're just using the .Net framework class library (and of course any Smalltalk classes). Where it gets hairy is if you try to bring in an outside .Net library or do COM work through .Net. Then you have to use .Net's own late-binding mechanisms which are a pain. In Smalltalk you get late binding for free. With .Net you have to do gymnastics. It supports it, but it's no picnic. One of the projects I've had in the back of my head for a while is maybe I'll try my hand at improving the bridge for these scenarios. The experience should be better.

Another option I heard about a couple years ago was WxWindows for Squeak. I've heard complaints about it, because it makes you do some gymnastics, but I think it works.

Re: Back to the future: Smalltalk

Mark Miller — Fri, 23 May 2008 04:08:24 -0000

Thanks for your response. The "day job" thing was bugging me, because I wondered if other programmers were seeing something I wasn't. I haven't really had any work experience with it, so I wondered if in "the real world" there were difficulties with it that I haven't seen yet, and what those specifically were. I've heard from people who have used Smalltalk in work settings. Some have praised it. Some have said bad things about it. Ironically, I've found the people who didn't like it didn't understand what it was really about. It made me wonder how they got hired to do it in the first place. It felt like I was talking to someone who used C but didn't understand how pointers worked.

I am familiar with the ESUG list of companies. Things like this give me hope. :)

I consider myself lucky that I was exposed to Smalltalk in college. It was part of a senior course on programming languages, exploring how they worked; different types of systems (compiled, interpreted), different types of languages (imperative, functional, object-oriented). It was a long time ago, so I don't remember exactly, but I think we covered Smalltalk for 2 weeks. I also got exposed to Lisp, in a different class, covering it for the same length of time. So I got a "taste" of them. Honestly, I think there are more jobs for Lisp programmers now than there were back then. ITA Software uses it, as does Orbitz. Both are in the U.S., I think.

As for what Paul Graham said, I'd like to believe that. The thing is, I think so many of the best developers don't believe they can find work using an esoteric language, so they learn Java or something, and apply their talents towards that. It's a self-fulfilling prophecy. A significant minority of talented developers are using and pushing for Ruby and Python to gain wider acceptance. That's good, IMO. They're a step in the right direction.

I have a suspicion what Graham said is boosterism. He's very much into believing that everyone should do what he's done with his life. I tend to agree with what he says, but I think if he were right then we'd see more development in esoteric languages than we do today.

I read up on what Graham wrote about his ViaWeb company. He began it using Lisp. He sold it to Yahoo. It's now called Yahoo Store. Yahoo eventually ported most of it to C++ and Perl. The reason they gave is they couldn't find enough Lisp developers to maintain it. Graham disagreed with that decision, of course, but that was Yahoo's determination.

I think Yahoo didn't have enough faith in Graham's decision to use Lisp. They probably said to themselves, "Yeah, whatever. He's a bit eccentric." If they understood his reasons, they would've pushed more for Lisp competence, pressuring CS programs to teach it more because, "We need Lisp developers." That's my perception of how these sorts of decisions happen at universities. They try to deny that they respond to industry pressure, but they do.

The thing is universities (in the U.S.) tend to teach them with a very clear message that "These are research languages. You won't be using them for your work." One of the things that's bugged me for a while is how Lisp was pigeon-holed as an AI language right from the start, as if it wasn't good for anything else. This doesn't do it justice.

Yes, startups have more freedom. So long as the founders can take care of most of the development themselves they can get away with that. I don't think corporate perception that there are too few developers who understand the esoteric languages is wrong. I think it would be a challenge to hire a whole lot of Lisp or Smalltalk developers. In my view it's a chicken and egg problem. You have on the one hand perception that there aren't enough developers who understand this stuff (so startups typically use PHP), and on the other you have a perception on the part of developers that they can't find work using them, so they go to PHP or Java.

I think there is a perception as well that in order to be able to handle a project you MUST be able to hire X number of developers at a rate range Y. In addition, a lot of commercial concerns look at industry support. There's a perception that you need good 3rd party support for whatever development platform you use, and if it doesn't have it it's not worth using. What Graham says, and I believe it's true, is if you use an esoteric language you don't need as many developers. In fact large teams are inefficient anyway. Secondly, you don't necessarily need the third party support if you have a good technical team that can roll its own solutions. So it takes someone brave, willing to buck convention, to do this. They also have to believe that their technology is a competitive advantage. Unfortunately in the U.S. for the past several years there's been a growing belief that "I.T. doesn't matter" (ie. technology is not a differentiator), and that IT management is what can differentiate one company from the next. There's some truth to that, but it's not the whole story.

Re: Switched to Mephisto

margesh — Wed, 21 May 2008 09:19:45 -0000

Mephisto is best blogging system ever

Re: Back to the future: Smalltalk

jozilla — Mon, 19 May 2008 06:24:51 -0000

The fact that you can rarely use Smalltalk in a day job has more to do with Smalltalk's (lack of) popularity than with its intrinsic qualities. As far as I know, there are only a handful of companies in Belgium that use Smalltalk, let alone offer Smalltalk jobs. The most well known is MediaGeniX. It seems they don't even require experience with Smalltalk anymore. In one of their job postings they say that applicants will be trained in the Smalltalk development environment. The European Smalltalk Users Group (ESUG) has an extensive list of companies that use Smalltalk.

Lisp has the same problem. I found a list of Lisp companies, with only one from Belgium: PEPITe, a spin-off company of the University of Liège.

So I didn't say that Smalltalk is not suitable for real-world software development, but it is just not used much in industry. As far as I know, most of the traditional programming jobs require either Java, .NET or C++. Furthermore, most programmers have no experience with Smalltalk. To the best of my knowledge, the only university in Belgium that actively teaches Smalltalk is the Vrije Universiteit Brussel, who perform research in programming languages. They host the Planet Smalltalk blog, hosted ESUG meetings and have employed some well-known people in the Smalltalk community (e.g. Roel Wuyts).

I can imagine that managers are afraid that they won't find programmers to maintain their codebase if they use Smalltalk. Although Paul Graham once said: "if a company chooses to write its software in a comparatively esoteric language, they'll be able to hire better programmers, because they'll attract only those who cared enough to learn it", I have a slight feeling that most companies don't think that way :-)

Of course, it all depends on the career you choose (and what you like to do). I can imagine that startup companies have more freedom to pick their preferred development tools.

Re: Back to the future: Smalltalk

jozilla — Sun, 18 May 2008 09:50:52 -0000

Thanks for pointing this out!

Re: Back to the future: Smalltalk

jozilla — Sun, 18 May 2008 09:42:53 -0000

Thanks for your kind words! I'll update the blog post with a few lines on Smalltalk being written in itself.

From reading Jeffrey Massung's blog, Dolphin Smalltalk indeed seemed to offer easy integration of Windows libraries (e.g. DirectX). I am not sure how Squeak compares to this. They have an FFI interface and a .NET bridge as far as I know. An advantage of Squeak to me is that it runs on many platforms, including mobile devices.

Re: Back to the future: Smalltalk

Nick Sieger — Fri, 16 May 2008 21:27:07 -0000

The swing example you followed must be old. JRuby does allow you to pass a block as event listeners. See http://www.infoq.com/articles/jruby-deployment-... for example.

Re: Back to the future: Smalltalk

Mark Miller — Fri, 16 May 2008 12:50:25 -0000

Help me understand something. I remember listening to a podcast with a guy (can't remember the name) talking about "aggressive learning" (I think). He said he gets into learning all sorts of languages, taking what he's learned and using it with the popular stuff. He used this same phrase you used when talking about the state of Smalltalk in the 1990s. He said developers used to say back then, "This is great, but I can't use it at my day job." What does that mean exactly? He said this in conjunction with also saying something to the effect that Smalltalk was an "ivory tower" language, used in university study programs but not much else. There are some cases in point where this obviously doesn't apply. Smalltalk continues to be used at some major firms. Squeak is getting used in a few startups successfully.

I can understand the obvious that it's not the same environment, computing model, or language that a company might already be using, but did it also mean that anyone who used it would effectively be isolated and unable to work in groups by virtue of the fact that everyone else didn't understand it, and refused to do so? Was it also cultural?

In the case of Squeak it's able to interoperate with C. I've seen some Squeak libraries that use C-based DLLs. I haven't researched how it's done, but I know it works, and it seems to be something built into the language/VM.

Maybe the problem in the 90s was that commercial Smalltalks (which I think were pretty much all that existed, with the exception of GNU Smalltalk) didn't interoperate with other stuff well, if at all. From my reading of Smalltalk's history, the fact that the most popular implementation was commercial was a big part of the problem.

I remember a time when Smalltalk was actually popular. When I was in college, in the 1990-1992 timeframe, I used to see want ads for Smalltalk programmers on a fairly regular basis. The demand for it was not nearly as high as for C, but from what I remember it seemed like it was more popular than Ruby on Rails is today. By the time I graduated in 1993, though, the demand for Smalltalk had disappeared. I noticed this and wondered why it happened. I had been introduced to Smalltalk very briefly in my 4th year of college. I really enjoyed it. I was beginning to think that maybe I could find work using it, but those hopes were dashed pretty quickly.

There are some first-hand account blog posts, and even some information pages around that tell the tale about what really happened to Smalltalk in the 1990s. It reads like a tragedy. It sounds like corporate mismanagement brought it down. When the corporate entities died, Smalltalk "went down with the ship". It didn't totally disappear, but it might as well have. Cincom managed to salvage the leading commercial implementation, VisualWorks. They're selling it as ObjectStudio today.

I think that Squeak offers Smalltalk a second chance at life, since it's not tied to any corporate entity. From what I've read GNU Smalltalk is kind of a weak implementation of the system. It's best used for scripting. Squeak is a full implementation of Smalltalk-80, and then some. Seaside has been breathing some life into Smalltalk. It hasn't taken off like RoR, but it's gotten some interest.

Re: Back to the future: Smalltalk

Mark Miller — Fri, 16 May 2008 12:16:24 -0000

Dolphin Smalltalk was discontinued last year, unfortunately. I guess the Community Edition is still around.

Re: Back to the future: Smalltalk

scott lewis — Fri, 16 May 2008 00:01:20 -0000

It's so great to read something positive about Smalltalk. You've really done a great job of explaining why it works so well. I think the only thing I would emphasize additionally is that Smalltalk is mostly written in Smalltalk. So when you subclass any object, you can go back up the chain of inherited objects and see how everything works. Likewise when you hit an error/bug, the debugger lets you delve about as deeply as you could possibly want into what is going wrong, and why it is an error.

One of the most subtle tests for objection-orientation for me is whether one uses "case" statements in code. In the most pure O-O they really are not needed, as you instead forward the message to the object in question, and it should "know" what to do.

In the general C family, it would seem Objective-C comes closest to Smalltalk, and it's in wide use in the Mac's Xcode. You can get around static typing there by using the "id" type -- which drives conventional C programmers up the wall, but there you go.

Of course, there are a lot of times when the rigid object model of Smalltalk can be a little frustrating, and one begins to wonder if multiple inheritance might work better. In general, I don't think so, but there is an interesting approach to this suggested by the language Self, which is prototype-based. Self allows the manner of inheritance itself to be specified (in a sense) which creates one of those great complexity from simplicity situations -- just like O-O itself.

If you like Squeak, you should really also take a look at Dolphin Smalltalk. It hooks directly into Windows in a great way, and is a really well developed, mature environment. There is a free Community Edition, and the Professional edition (I think it's around $400 -- been a while since I bought mine) allows you to compile executables.

Re: Back to the future: Smalltalk

jozilla — Thu, 15 May 2008 11:53:47 -0000

Indeed, Smalltalk wasn't meant to be the endpoint. A Conversation with Alan Kay is a great summary of his viewpoint on this.

Thanks for the insights on Erlang. I actually wonder how Croquet's TeaTime compares with Erlang's concurrency mechanisms (or E's for that matter).

It is indeed a shame that people easily give up on different languages or systems. I can understand the practical issues though. You can rarely employ these languages in a day job, and it can be hard to cooperate with others who program in traditional languages. However, it all depends on what your goals are. I believe that in research one should be able to experiment and select the right tool for the job.

Re: Back to the future: Smalltalk

Mark Miller — Thu, 15 May 2008 05:33:19 -0000

I think you are on the right track with your analysis of the situation. If you read up/watch some of what Alan Kay has said about OOP you'll see that he considered it as just a starting point. In fact, he's said he's apologized for years that he coined the term, because it got everyone focused on objects, rather than the messages. OOP wasn't supposed to be the end all, and be all of programming. He expected that a younger generation of programmers (in the 1980s and thereafter) would come along and improve upon, but he says that didn't happen.

I haven't looked at Erlang, but I've done a little research on it. It takes a similar "no centers" approach to development, and uses late binding, as Smalltalk does. It also uses message passing, but it's asynchronous. Instead of objects, it has things called "nodes", I believe. Ericsson invented it, and has been using it for years to run its communication servers. From what I've heard they haven't had a server crash yet. Even if they run into an error, it's like Smalltalk. A developer can just fix the error on the spot, and the process continues.

I think the reason we don't see more development using systems like this is not because they don't work, but because most programmers don't have the background to deal with them. Give Erlang to most developers and they'll give up on it in short order. They won't understand it. The same thing has happened with Smalltalk. People take one look at it and say, "Hey, why can't I use my code editor and my version control system with it? Why does it insist that I only use one language?", and they give up on it. Quite a shame.

Re: OneNote: a hidden Microsoft Office gem

jozilla — Fri, 09 May 2008 05:00:29 -0000

Thanks! That's a great suggestion since it runs on multiple platforms. I just signed up for an invitation to the beta program :-)

Re: OneNote: a hidden Microsoft Office gem

Fri, 09 May 2008 01:08:45 -0000

Ever heard about Evernote? http://www.evernote.com/about/what_is_en/
Might be a nice addition.

Re: Ubuntu running Windows in VMware Player

jozilla — Tue, 06 May 2008 12:41:58 -0000

You're welcome, good luck!

Re: Ubuntu running Windows in VMware Player

Manuel — Tue, 06 May 2008 11:58:24 -0000

Thanks for the heads up. I actually have no issues with USB--using Mandriva 2007.1 here. It's just the emulator that won't start. I'll install VS 2005 and see where I go from there.

Regards,

Manuel