I've posted here an on my site why I chose to use C++. Since you're so motivated to waste my time by annoying me by porting Micropolis to another language (and you're really sure it will annoy me), then go right ahead! Get started! What are you waiting for? Stop wasting your time on slashdot! Please report back when you're done, and I will pretend to be annoyed, if that is what truly motivates you.
For inspiration and ideas, you might want to read some of the stuff Alan Kay and I were discussing about opening up and programming SimCity with visual languages, ala Robot Odyssey, on my blog.
The UI is quite abstract. I had to cleanly separate the UI and the simulator engine to port SimCity to NeWS, and TCL/Tk maintained that separation. The new MicropolisCore has all the UI code taken out, so all the rendering and event handing has to be implemented outside of the simulator core, in a scripting language like Python, with a graphics library like Cairo, and a user interface toolkit like GTK.
The new Micropolis core simulator engine talks to the tile engine through efficient raw memory pointers. Micropolis hands the tile engine a pointer to its tile memory and a description of the format, and a tile bitmap and a description of the tile layout (including an indirect tile map). The tile engine uses Cairo to render the tiles. Cairo is a much more advanced imaging library than SDL, and most importantly it supports Unicode and fully internationalized text via the Pango library (OLPC activities must support many different languages). Cairo enables scaling and transformation, vector graphics, anti-aliasing, clipping, SVG, etc. It is more like the PostScript based user interface to SimCity I developed for NeWS. The Cairo tile engine supports zooming into the city, and the old NeWS version used to let you play the game at any scale. Zooming is very important on the OLPC's 200 dot per inch resolution screen, because the unscaled tiles are very small, and kids like to zoom in to look closer at the city, so it's easier to see. Another reason for using Cairo is that it's great for drawing map overlays and data visualizations, with scalable vector graphics, transparency, etc. And also, Cairo is hardware accelerated on Linux, the Mac and Windows, so it can draw extremely fast and still look pretty.
In order to modify a C program to use structures and function pointers instead of global variables, you have to meticulously modify almost every line of code. And that is simply far too much work to do, and would be extremely tedious and error prone.
The one thing C++ is really great at, is taking a C program with global variables and global functions, and easily transforming it into a C++ class that encapsulates all the variables and functions, without disrupting every line of code. Because "this" is an implicit argument. Because member references don't have to explicitly go through "this".
But if you did that kind of a transformation on C code by hand, you would have to rewrite every function signature to take an explicit "this" argument, and rewrite every variable or function reference to go through the "this" pointer.
Actually, I really dislike GTK's Python binding mechanism. It's horribly complex, totally undocumented, and very brittle. It doesn't handle other languages, either. SWIG is a much more advanced, much easier to use tool. I also don't like GTK's "yet another object system". It accounts for most of the time spent by OLPC Sugar python applications initializing. It takes a horribly long time for GTK to initialize, and it's not Python's fault, it's GObject's fault. There's nothing worse than having several object systems, especially when some of them are slow, non-standard and difficult to work with.
Have you ever looked at or used the COM macros for generating C++ vtable layouts? It's attrocious! Actually, I really like COM for what it is (I use XPCOM at work, since we use xulrunner to implement TomTom Home), but COM's C bindings are total crap. It's much easier to use from C++ with templates, ala ATL (ActiveX Template Library), WFC (Windows Foundation Classes), etc. The MFC COM and OLE stuff is much worse, but not as bad as the C stuff.
C++ templates can't hold a candle to Common Lisp macros. C++ templates are a totally different animal, totally inferior to the macros that Lisp has had for many years. The arcane C++ syntax makes it impossible to support the kind of high level metaprogramming macros that Lisp so easily supports.
Agreed. But Java isn't exactly what I'd call a scripting language. What I expect from a scripting language is the easy ability to interact with it on a command line or interactive interpretive interface (like Smalltalk). (It doesn't actually have to be interpreted, just compiled fast enough that I don't notice, like many Lisps and OpenLaszlo implement their interactive command line.) Java's problem is not just that the compilers and tools aren't set up for interactive evaluation, but also that the syntax makes it extremely tedious and required typing a lot of boilerplate structural crap. Of course there are some good interactive interpreted languages like JavaScript (Rhino), Python (Jython) and Groovy implemented in Java.
The great thing about integrating C++ modules like micropolis with Python is that you can play around with it from the keyboard, test and debug it easily, and get it to work quickly. Immediately after I compiled the micropolis module for the first time, I was able to import it, make a simulator, load a city, set the tax rate, and run the simulation loop, all from the keyboard! Then I copied that code I typed into a python file, and I had the first test framework.
Yes, it'll be great to port Micropolis (SimCity) to embedded devices!
I'd like to port it to the TomTom, and hook it up so you can operate the bulldozer or road building tool via GPS, just by driving around!
-Don
(Disclaimer: I work for TomTom, and this should NOT be interpreted as a product announcement!)
Other people wanted to use SimCity for the less noble goal of teaching people what to think, instead of just teaching them to think.
Everyone notices the obvious built-in political bias, whatever that is. But everyone sees it from a different perspective, so nobody agrees what its real political agenda actually is. I don't think it's all that important, since SimCity's political agenda pales in comparison to the political agenda in the eye of the beholder.
Some muckety-muck architecture magazine was interviewing Will Wright about SimCity, and they asked him a question something like "which ontological urban paridigm most influenced your design of the simulator, the Exo-Hamiltonian Pattern Language Movement, or the Intra-Urban Deconstructionist Sub-Culture Hypothesis?" He replied, "I just kind of optimized for game play."
NeWS was like AJAX, but with PostScript instead of JavaScript for programming, with PostScript instead of DHTML for rendering, and with PostScript instead of XML for data representation.
You can do anything in C you can do in C++, but it's a total pain in the ass. Why, you can even write a C++ compiler in C that compiles C++ into C, and call it CFront.
If you want to see what happens when you try to simulate C++ interfaces in C, then look at Microsoft's COM header files, which do exactly that.
The reasons I translated the C code to C++ were: 1) eliminate all global variables, so multiple simulators could exist simultaneously without interfering with each other, 2) define all interfaces in one place so it's easier to work with and evolve the code, 3) enable SWIG to automatically generate an object oriented wrapper for any of a large number of scripting languages, like Python, 4) Impose some sane programming conventions on the code, for reliability and readability's sake.
I hate C++ as much as anyone else does, probably more than most. But I know when it's better to use C++ than C, and this is one of those times.
-Don
``The more important question of whether or not systems like C++ and X-Windows cause actual permanent brain damage in humans, will have to wait for the autopsies of human brain doners. That won't be soon, since they're still only in the animal experimentation stages. But I think it's reprehensible how they're exposing all those poor innocent bunnies to shoddy static class libraries without garbage collection, and high doses of raw Motif.''
[From comp.lang.functional, ``Re: Effect of programming languages on thought'', Don Hopkins]
The original SimCity code written in C ran just fine of an 8 bit 1.02 MHz 6510. And I've optimized to run even more efficiently since then. So worrying about Python slowing SimCity down is totally lacking in perspective -- penny wise but pound foolish. SimCity is already many orders of magnitude faster than it needs to be. Anyway, the core simulator is written in C, so Python doesn't slow it down at all. You should learn more about Python programming, developing Python modules in C and C++, and using SWIG for integrating Python and native code, and using Python as an embedded application extension language, before "making such [...] comments".
Using Python drastically speeds up the software development process, which is a great thing when software developer's time is so expensive, and computers are so fast. Python is also is a much easier language for kids to read, learn and program -- and the OLPC is an educational project, not a laptop project.
To clear up some confusion: The old version of SimCity/Micropolis uses TCL/Tk to implement the user interface. That is the version currently being distributed on the OLPC. The new version is rewritten in C++, and has all the TCL/Tk and user interface stuff ripped out of it. I converted it to C++ for the reasons I described in other posts (quoted above), so it is now modular and can be run through SWIG to integrate it with many different scripting languages.
But the core simulator is independent of Python, and runs extremely fast (the TCL/Tk version can run more than a year a second on the relatively slow OLPC). The OLPC uses Python as its standard system wide programming language, and all of its important libraries (like Cairo and Pango) are integrated as Python modules. So it makes the most sense to use SWIG to cast Micropolis into a Python module, first. Of course SWIG also makes it easy to integrate it with any other scripting language.
If it's not immediately obvious to you (or even if it is), why anyone would want to integrate SimCity with a scripting language, instead of just writing the whole thing in C, then you should read some of the discussions I've been having with Alan Key about that topic, on my blog.
TomTom's "MapShare" technology lets users correct maps and generate their own content.
You can actually correct the problems on the TomTom device when you encounter them, so they're applied to your map immediately, and they're uploaded and shared with other users when you hot-sync your TomTom with your PC.
Of course you can also download other user's corrections. You can choose to only use corrections you made yourself, or download corrections verified by TomTom, corrections to POIs you subscribe to, corrections from trusted sources, corrections reported by many people, or corrections reported by some people.
Of course (as demonstrated in the following video), sometimes a goof-ball gets ahold of the cheat code, and messes everything up for everybody, so you may not want to download any change that haven't been reported by many people!
http://www.youtube.com/watch?v=GU2iQX4vJ10
-Don
I've just posted a whole bunch of stuff about OLPC SimCity, including some exciting discussion with Alan Kay about eToys, Robot Odyssey, Visual Programming, and teaching kids to program, to my blog:
SimCity for OLPC (One Laptop Per Child): Applying Papert's Ideas About Constructionist Education and Teaching Kids to Program:
http://www.donhopkins.com/drupal/node/129
The rules are written in C, which I have ansified and cleaned up a lot, but not fundamentally changed.
The first step would be to write a roadmap and documentation, which I've started in a skeletal form.
One necessary step is to convert the C code to C++ so it's possible to define clean interfaces between software modules, and make multiple instances of the simulator that don't interfere with each other, as well as easily interfacing it to Python using the SWIG interface generator.
That should be done in a language-neutral way, so you could plug the simulator engine into many different languages and programming systems.
Then more work needs to be done to open it up, and make it re-vectorable (plug-ins, events, callbacks, hooks, aspect oriented programming, etc), so you can replace and extend the various modules with the host language(s), eventually re-implementing most if not all of SimCity in another language.
GUI & OLPC support designed and implemented by Don Hopkins.
Developed thanks to the generous support of:
Maxis
Electronic Arts
DUX Software
Sun Microsystems
Turing Institute
Kaleida Labs
OLPC
John Gilmore
Internet Archive
TomTom
Special thanks to:
Will Wright
Jeff Braun
Mike Perry
Charles Norman
John Riccitiello
Steve Seabolt
Amanda Taggart
JoAnn Covington
Brian Rubin
Alicia Truby
Jim Mackraz
Jamie Doornbos
Eric Bowman
Arthur van Hoff
Dug Scoular
Bob Adams
John Ousterhout
John Gilmore
Seymour Papert
Alan Kay
Mark Weiser
Ben Shneiderman
Walter Bender
SJ Klein
Mary Lou Jepsen
Ted Selker
Mitch Bradley
Jim Gettys
Justin McCormick
Cassidy Wright
Slats
Broken Robot
Professor Johnson
Kitty Puff Puff
Misty the WonderBot
Super ChiBot
Chia's Little Helper
(Robot Action
Combat Cluster
MultiBot)
I wonder if it would be possible for you to orthogonalize the "world environment objects" from the control environment drivers so that a variety of people could try to make "interactive brains" for SimCity. (Perhaps on the OLPC, the DBus could be used to connect the two.) This would allow many different ideas to be explored for how different ways and different ages of children could control this game.
For example, I have a grad student (Alex Warth, a fulltime researcher at Viewpoints) who is the best I've worked with in at least 20 years. He has been implementing many languages with a deep meta system he has built. I would love to get him started on trying one version of a language and interface for children to see and put in new rules. It would be great if others did this also. These control modules could be either loaded with SimCity, or more safely, could be in a separate address space and act as an intermediary between users and the rest of the game.
It would be terrific if others tried to do their own modules, just as you stated in your plans. A little friendly design and implementation rivalry would be great for the kids (and we would all learn a lot also). (Do you know of Mike Genesereth's Game Competitions that he runs from a Stanford website? He basically supplies game environments and grad students around the world write programs that try to learn and play the game against each other. Similarly, SimCity is a very rich game environment and could attract lots of very smart designers to make it more extensible and malleable.
To get Alex started, we would need to articulate and list the "rules, heuristics, etc." that the classic SimCity uses. I don't know what form these are in, but I'll bet they look nicer in the reimplementations you've done than in the original game. You probably understand them better than anyone. In any case, if you are interested in this, Alex and I could come visit you (where are you?) and try to pick your brains about ways to give form to these rules that make sense in the world of relatively young children.
I'm writing to applaud you for your plans to reimplement SimCity for children on the OLPC.
My main complaint about this game has always been the rigidity, and sometimes stupidity, of its assumptions (counter crime with more police stations) and the opaqueness of its mechanism (children can't find out what its actual assumptions are, see what they look like, or change them to try other systems dynamics).
So I have used SimCity as an example of an anti-ed environment despite all the awards it has won. It's kind of an air-guitar environment.
In the past, I tried to get Maxis to take the actual (great) educational possibilities more seriously, but to no avail.
Going to Python can help a few areas of this, but a better abstraction for the heart of Sim-City would be a way to show its rules/heuristics in a readable and writable form. Both of these could be stylized to put them in the child's own thinking and doing world. For example, just the simple route of making a drag and drop scripting interface for Etoys allows children to make very readable and writeable scripts and helps the children concentrate on what they are trying to do. A carefully designed object system (that is filtered fro children) can expose the environment so they can really think about it.
I'm not at all suggesting that Etoys be used here, but I am suggesting that some deep design be done to come up with a "behavior modification interface" that allows real creativity on the part of the children. So it is much more than stringing black boxes together or having to deal with fragile procedurals.
I sense that you have some interests in making SimCity really a microworld for children's learning and exploration from reading your webpage.
Children in 4th - 6th grade can do a lot here if they are given a good UI and tools. So, we could think of part of this project as a "pre-Python" UI.
Scalability and non-scalability of ideas are interesting. Rocky's Boots is still one of the best ever games that provide profound learning experiences. The extension of this to Robot Odyssey didn't work because the logic and wires programming didn't scale well enough -- the bang per effort dropped off precipitously. I was Chief Scientist at Atari at that time (Warren Robbinet worked for me) and I worked with TLC to try to get them to realize that something like Logo, or even better, a rule-based robot programming system, was needed. The failure of Robot Odyssey really pained me because I thought that the concept of this game was one of the best ever (still is). But it just needed a much better notion of how the children were going to program the robots. I think the same goes for SimCity.
I like the idea of forcing people to stay married, as long as you let gays get married and force them to stay that way too. Let's start with the current crop of Republican presidential candidates. Disqualify all the ones who've been divorced, and all the ones who are secretly gay, and you wouldn't have any Republican presidential candidates left!
You are totally talking out of your ass, and you're afraid to post that tripe under your own name, because you know you don't have a leg to stand on, and can't support your side of the argument with verifiable facts. Why don't you sign into Slashdot under your real name, and post some references to back up the ridiculous claims you're making? Just because you say it but won't sign your name to it, doesn't mean it's true.
-Don
Re:Porting SimCity to Python + OLPC's "Sugar" gui
on
One SimCity Per Child
·
· Score: 2, Interesting
First, the legal details:
The GPL source code version of SimCity will not be called "SimCity", but we will use the SimCity source code to make a city building game called "Micropolis", which was the original working title of SimCity.
That's because EA reserves the right to review and QA the official version of the game that's published under the name "SimCity" on the OLPC.
So we can make improvements to the TCL/Tk version of Micropolis (based on the GPL source code), and submit them to EA for review and QA, which if they approve, will be used as the officially branded version of SimCity for the OLPC.
It will be the same code, but the only difference is the name, which EA understandably wants to protect, be ensuring the quality and integrity of OLPC SimCity.
That said, I strongly encourage you to improve the TCL/Tk version of Micropolis, please!
I think it would be great to keep the TCL/Tk development line of the code alive and strong, and modernize and improve it!
It will be a fair amount of work to upgrade it, but that will be a lot less work and time than recasting SimCity in Python will require.
Of course both projects can go on in parallel, learning from each other.
Since the Python version won't be ready for a long time, we can submit an upgraded version of TCL/Tk Micropolis as the official version of OLPC SimCity, based on the latest TCL/Tk, including further OLPC/Sugar integration \(like making sound work better, supporting the gamepad, etc).
TCL/Tk is a very cleanly written language and well designed user interface toolkit, which has grown and improved a lot since I took a snapshot of the code from around 1993 to use for SimCity.
I have found memories of meeting Professor Ousterhout in his office on the UC Berkeley campus, and showing him the TCL/Tk version of SimCity, and thanking him for the great work he'd done and shared for free.
The reason I want to convert the TCL/Tk version of SimCity to Python is because it's the "official" programming language for OLPC Sugar applications.
In the short term, the OLPC will include applications like SimCity that uses TCL/Tk, eToys that uses Smalltalk/Squeak, etc.
But in the long term, new software developed for the OLPC should be written in Python, to save memory by sharing modules, and to use great modules like Cairo for scalable stencil/paint graphics, Pango for rendering internationalized text, and all the Sugar modules written in Python.
Despite not being written in Python, eToys and SimCity are included with the OLPC.
The point is to educate and inspire people to develop new and better software, to bootstrap the software of the future.
I'm compelled to move forward and develop a new visual programming language, better and easier to use than Python or TCL, that kids can use to reprogram the behavior and create extensions to SimCity, and even write their own games and who knows what else.
-Don
Alan Kay wrote this on the OLPC Sugar mailing list, about implementing visual programming languages like eToys in Python:
Guido knows that I've been advocating that the Python folks should do Etoys or a very Etoys like environment in Python (and that the rest of the OLPC be given an objectification and media and scripting integration that is Etoys like).
However, there are simply zillions of things left to be done everywhere for OLPC so the first round of SW on the XO will be more of a gathering of "suggestive" features and abilities (of which Etoys is just one). That seems fine to me.
Viewpoints Research (our little non-profit) doesn't have any "ego or identity" staked around whether the children's authoring environment is Python based or Squeak based. I have said many times that, if the general integrative base of XO is to be Python, then the Etoys-like authoring should be based in Python also.
However, I will personally fight to the death to make sure that there is a children's authoring environment that allows even young children to do simulation style programming with very ri
Well if you're not a Christian conservative, then do you support Bush and Cheney who pander to them? Do you believe they're only lying to the gullible Christians in order to get votes? Or do you think they believe what they say? In either case you're an idiot to support Bush and Cheney if you disagree with the religious views they claim, but you're not a Christian.
There is no original article. The onslaught of hyperlinks was the article!
What's wrong with links? You don't HAVE to click on them. Or does your OCD torment you over every link left unclicked?
-Don
Forth on the OLPC, OpenFirmware and ChipWits
on
One SimCity Per Child
·
· Score: 2, Interesting
Hey, why don't you help Mitch Bradley port his excellent Forth system to run on the OLPC? It already does, but it's in the OpenFirmware boot ROMs, so it runs before Linux even boots. But the same Forth system runs quite nicely under Unix as well (without all the direct hardware access:), and is quite luxurous. I used it on the Sun 3 and Sparcstation years ago, and worked at Sun as his summer intern on CForth, another portable Forth system. Mitch is one of the best and most accomplished Forth programmers on or off the planet. If you want to do Forth on the OLPC, Mitch Bradley is definitely the dude to talk to about it!
Sure you're serious. And you probably take Stephen Colbert at face value.
You conservatives are so funny, but totally out of touch with reality.
You impeached Clinton for a blow job, but then you defend Bush no matter what he illegal and immoral things he does.
And you all call yourselves Christians. Ha.
Everyone notices the obvious built-in political bias, whatever that is. But everyone sees it from a different perspective, so nobody agrees what its real political agenda actually is. I don't think it's all that important, since SimCity's political agenda pales in comparison to the political agenda in the eye of the beholder.
Some muckety-muck architecture magazine was interviewing Will Wright about SimCity, and they asked him a question something like "which ontological urban paridigm most influenced your design of the simulator, the Exo-Hamiltonian Pattern Language Movement, or the Intra-Urban Deconstructionist Sub-Culture Hypothesis?" He replied, "I just kind of optimized for game play."
Slashdot Mirror
By "understanding rhetoric" does that mean "understanding that you don't mean what you say"?
-Don
I've posted here an on my site why I chose to use C++. Since you're so motivated to waste my time by annoying me by porting Micropolis to another language (and you're really sure it will annoy me), then go right ahead! Get started! What are you waiting for? Stop wasting your time on slashdot! Please report back when you're done, and I will pretend to be annoyed, if that is what truly motivates you.
For inspiration and ideas, you might want to read some of the stuff Alan Kay and I were discussing about opening up and programming SimCity with visual languages, ala Robot Odyssey, on my blog.
-Don
The UI is quite abstract. I had to cleanly separate the UI and the simulator engine to port SimCity to NeWS, and TCL/Tk maintained that separation. The new MicropolisCore has all the UI code taken out, so all the rendering and event handing has to be implemented outside of the simulator core, in a scripting language like Python, with a graphics library like Cairo, and a user interface toolkit like GTK.
The new Micropolis core simulator engine talks to the tile engine through efficient raw memory pointers. Micropolis hands the tile engine a pointer to its tile memory and a description of the format, and a tile bitmap and a description of the tile layout (including an indirect tile map). The tile engine uses Cairo to render the tiles. Cairo is a much more advanced imaging library than SDL, and most importantly it supports Unicode and fully internationalized text via the Pango library (OLPC activities must support many different languages). Cairo enables scaling and transformation, vector graphics, anti-aliasing, clipping, SVG, etc. It is more like the PostScript based user interface to SimCity I developed for NeWS. The Cairo tile engine supports zooming into the city, and the old NeWS version used to let you play the game at any scale. Zooming is very important on the OLPC's 200 dot per inch resolution screen, because the unscaled tiles are very small, and kids like to zoom in to look closer at the city, so it's easier to see. Another reason for using Cairo is that it's great for drawing map overlays and data visualizations, with scalable vector graphics, transparency, etc. And also, Cairo is hardware accelerated on Linux, the Mac and Windows, so it can draw extremely fast and still look pretty.
-Don
In order to modify a C program to use structures and function pointers instead of global variables, you have to meticulously modify almost every line of code. And that is simply far too much work to do, and would be extremely tedious and error prone.
The one thing C++ is really great at, is taking a C program with global variables and global functions, and easily transforming it into a C++ class that encapsulates all the variables and functions, without disrupting every line of code. Because "this" is an implicit argument. Because member references don't have to explicitly go through "this".
But if you did that kind of a transformation on C code by hand, you would have to rewrite every function signature to take an explicit "this" argument, and rewrite every variable or function reference to go through the "this" pointer.
Actually, I really dislike GTK's Python binding mechanism. It's horribly complex, totally undocumented, and very brittle. It doesn't handle other languages, either. SWIG is a much more advanced, much easier to use tool. I also don't like GTK's "yet another object system". It accounts for most of the time spent by OLPC Sugar python applications initializing. It takes a horribly long time for GTK to initialize, and it's not Python's fault, it's GObject's fault. There's nothing worse than having several object systems, especially when some of them are slow, non-standard and difficult to work with.
Have you ever looked at or used the COM macros for generating C++ vtable layouts? It's attrocious! Actually, I really like COM for what it is (I use XPCOM at work, since we use xulrunner to implement TomTom Home), but COM's C bindings are total crap. It's much easier to use from C++ with templates, ala ATL (ActiveX Template Library), WFC (Windows Foundation Classes), etc. The MFC COM and OLE stuff is much worse, but not as bad as the C stuff.
C++ templates can't hold a candle to Common Lisp macros. C++ templates are a totally different animal, totally inferior to the macros that Lisp has had for many years. The arcane C++ syntax makes it impossible to support the kind of high level metaprogramming macros that Lisp so easily supports.
-Don
Agreed. But Java isn't exactly what I'd call a scripting language. What I expect from a scripting language is the easy ability to interact with it on a command line or interactive interpretive interface (like Smalltalk). (It doesn't actually have to be interpreted, just compiled fast enough that I don't notice, like many Lisps and OpenLaszlo implement their interactive command line.) Java's problem is not just that the compilers and tools aren't set up for interactive evaluation, but also that the syntax makes it extremely tedious and required typing a lot of boilerplate structural crap. Of course there are some good interactive interpreted languages like JavaScript (Rhino), Python (Jython) and Groovy implemented in Java.
The great thing about integrating C++ modules like micropolis with Python is that you can play around with it from the keyboard, test and debug it easily, and get it to work quickly. Immediately after I compiled the micropolis module for the first time, I was able to import it, make a simulator, load a city, set the tax rate, and run the simulation loop, all from the keyboard! Then I copied that code I typed into a python file, and I had the first test framework.
-Don
Yes, it'll be great to port Micropolis (SimCity) to embedded devices! I'd like to port it to the TomTom, and hook it up so you can operate the bulldozer or road building tool via GPS, just by driving around!
-Don
(Disclaimer: I work for TomTom, and this should NOT be interpreted as a product announcement!)
From Designing User Interfaces to Simulation Games, a summary of Will Wright's talk to Terry Winnograd's User Interface Class at Stanford, in 1996.:
-Don
NeWS was like AJAX, but with PostScript instead of JavaScript for programming, with PostScript instead of DHTML for rendering, and with PostScript instead of XML for data representation.
-Don
You can do anything in C you can do in C++, but it's a total pain in the ass. Why, you can even write a C++ compiler in C that compiles C++ into C, and call it CFront.
If you want to see what happens when you try to simulate C++ interfaces in C, then look at Microsoft's COM header files, which do exactly that.
The reasons I translated the C code to C++ were: 1) eliminate all global variables, so multiple simulators could exist simultaneously without interfering with each other, 2) define all interfaces in one place so it's easier to work with and evolve the code, 3) enable SWIG to automatically generate an object oriented wrapper for any of a large number of scripting languages, like Python, 4) Impose some sane programming conventions on the code, for reliability and readability's sake.
I hate C++ as much as anyone else does, probably more than most. But I know when it's better to use C++ than C, and this is one of those times.
-Don
The original SimCity code written in C ran just fine of an 8 bit 1.02 MHz 6510. And I've optimized to run even more efficiently since then. So worrying about Python slowing SimCity down is totally lacking in perspective -- penny wise but pound foolish. SimCity is already many orders of magnitude faster than it needs to be. Anyway, the core simulator is written in C, so Python doesn't slow it down at all. You should learn more about Python programming, developing Python modules in C and C++, and using SWIG for integrating Python and native code, and using Python as an embedded application extension language, before "making such [...] comments".
Using Python drastically speeds up the software development process, which is a great thing when software developer's time is so expensive, and computers are so fast. Python is also is a much easier language for kids to read, learn and program -- and the OLPC is an educational project, not a laptop project.
-Don
To clear up some confusion: The old version of SimCity/Micropolis uses TCL/Tk to implement the user interface. That is the version currently being distributed on the OLPC. The new version is rewritten in C++, and has all the TCL/Tk and user interface stuff ripped out of it. I converted it to C++ for the reasons I described in other posts (quoted above), so it is now modular and can be run through SWIG to integrate it with many different scripting languages.
But the core simulator is independent of Python, and runs extremely fast (the TCL/Tk version can run more than a year a second on the relatively slow OLPC). The OLPC uses Python as its standard system wide programming language, and all of its important libraries (like Cairo and Pango) are integrated as Python modules. So it makes the most sense to use SWIG to cast Micropolis into a Python module, first. Of course SWIG also makes it easy to integrate it with any other scripting language.
If it's not immediately obvious to you (or even if it is), why anyone would want to integrate SimCity with a scripting language, instead of just writing the whole thing in C, then you should read some of the discussions I've been having with Alan Key about that topic, on my blog.
-Don
TomTom's "MapShare" technology lets users correct maps and generate their own content. You can actually correct the problems on the TomTom device when you encounter them, so they're applied to your map immediately, and they're uploaded and shared with other users when you hot-sync your TomTom with your PC. Of course you can also download other user's corrections. You can choose to only use corrections you made yourself, or download corrections verified by TomTom, corrections to POIs you subscribe to, corrections from trusted sources, corrections reported by many people, or corrections reported by some people. Of course (as demonstrated in the following video), sometimes a goof-ball gets ahold of the cheat code, and messes everything up for everybody, so you may not want to download any change that haven't been reported by many people! http://www.youtube.com/watch?v=GU2iQX4vJ10 -Don
I've just posted a whole bunch of stuff about OLPC SimCity, including some exciting discussion with Alan Kay about eToys, Robot Odyssey, Visual Programming, and teaching kids to program, to my blog:
http://www.donhopkins.com/drupal
Here is the flurry of recent posts:
SimCity Rules: http://www.donhopkins.com/drupal/node/145
Python Plug-In Technologies for Extending OLPC SimCity: http://www.donhopkins.com/drupal/node/144
OLPC Visual Programming Languages for Education: http://www.donhopkins.com/drupal/node/143
Redesigning the SimCity User Interface for the OLPC: http://www.donhopkins.com/drupal/node/142
Ideas about OLPC SimCity GUI, Turtle Graphics, and Cellular Automata: http://www.donhopkins.com/drupal/node/141
Discussion with Alan Kay about Visual Programming: http://www.donhopkins.com/drupal/node/140
Discussion with Alan Kay about Robot Odyssey: http://www.donhopkins.com/drupal/node/139
SimCity, Robot Odyssey, and Visual Programming: http://www.donhopkins.com/drupal/node/138
OLPC Visual Programming Language Discussion with Guido van Rossum and Alan Kay: http://www.donhopkins.com/drupal/node/137
Slashdot OLPC SimCity Discussion: http://www.donhopkins.com/drupal/node/136
Responding to Alan Kay's criticisms of SimCity: http://www.donhopkins.com/drupal/node/135
Alan Kay's ideas about SimCity for OLPC: http://www.donhopkins.com/drupal/node/134
SJ Klein's OLPC Keynote at GDC Serious Games Summit: http://www.donhopkins.com/drupal/node/133
Alan Kay on Programming Languages: http://www.donhopkins.com/drupal/node/132
History and Future of OLPC SimCity / Micropolis: http://www.donhopkins.com/drupal/node/131
GPL Open Source Code of "OLPC SimCity" to be called "Micropolis": http://www.donhopkins.com/drupal/node/130
SimCity for OLPC (One Laptop Per Child): Applying Papert's Ideas About Constructionist Education and Teaching Kids to Program: http://www.donhopkins.com/drupal/node/129
Pie Menus for OLPC Sugar User Interface, in Python with GTK, Cairo and Pango modules: http://www.donhopkins.com/drupal/node/128
I replied to Alan Kay:
I'm thrilled to be working with you!
The rules are written in C, which I have ansified and cleaned up a lot, but not fundamentally changed. The first step would be to write a roadmap and documentation, which I've started in a skeletal form. One necessary step is to convert the C code to C++ so it's possible to define clean interfaces between software modules, and make multiple instances of the simulator that don't interfere with each other, as well as easily interfacing it to Python using the SWIG interface generator. That should be done in a language-neutral way, so you could plug the simulator engine into many different languages and programming systems. Then more work needs to be done to open it up, and make it re-vectorable (plug-ins, events, callbacks, hooks, aspect oriented programming, etc), so you can replace and extend the various modules with the host language(s), eventually re-implementing most if not all of SimCity in another language.
-Don
I recorded a demo of OLPC SimCity, and uploaded it to YouTube. Here is the URL, and the credits: http://www.youtube.com/watch?v=EpKhh10K-j0
OLPC SimCity
GPL Open Source Code
"Micropolis" Release
GUI & OLPC support designed and implemented by Don Hopkins.
Developed thanks to the generous support of:
Maxis
Electronic Arts
DUX Software
Sun Microsystems
Turing Institute
Kaleida Labs
OLPC
John Gilmore
Internet Archive
TomTom
Special thanks to:
Will Wright
Jeff Braun
Mike Perry
Charles Norman
John Riccitiello
Steve Seabolt
Amanda Taggart
JoAnn Covington
Brian Rubin
Alicia Truby
Jim Mackraz
Jamie Doornbos
Eric Bowman
Arthur van Hoff
Dug Scoular
Bob Adams
John Ousterhout
John Gilmore
Seymour Papert
Alan Kay
Mark Weiser
Ben Shneiderman
Walter Bender
SJ Klein
Mary Lou Jepsen
Ted Selker
Mitch Bradley
Jim Gettys
Justin McCormick
Cassidy Wright
Slats
Broken Robot
Professor Johnson
Kitty Puff Puff
Misty the WonderBot
Super ChiBot
Chia's Little Helper
(Robot Action
Combat Cluster
MultiBot)
I replied to Alan Kay:
I'm with you completely! Here are a couple messages I posted in the Slashdot discussion "One SimCity Per Child": http://games.slashdot.org/article.pl?sid=07/11/08/200234
Here's a great article, too!
http://arstechnica.com/news.ars/post/20071110-original-sim-city-donated-to-one-laptop-per-child-project.html
-Don
He replied:
Thanks very much Don!
A truly great opportunity here I think.
I wonder if it would be possible for you to orthogonalize the "world environment objects" from the control environment drivers so that a variety of people could try to make "interactive brains" for SimCity. (Perhaps on the OLPC, the DBus could be used to connect the two.) This would allow many different ideas to be explored for how different ways and different ages of children could control this game.
For example, I have a grad student (Alex Warth, a fulltime researcher at Viewpoints) who is the best I've worked with in at least 20 years. He has been implementing many languages with a deep meta system he has built. I would love to get him started on trying one version of a language and interface for children to see and put in new rules. It would be great if others did this also. These control modules could be either loaded with SimCity, or more safely, could be in a separate address space and act as an intermediary between users and the rest of the game.
It would be terrific if others tried to do their own modules, just as you stated in your plans. A little friendly design and implementation rivalry would be great for the kids (and we would all learn a lot also). (Do you know of Mike Genesereth's Game Competitions that he runs from a Stanford website? He basically supplies game environments and grad students around the world write programs that try to learn and play the game against each other. Similarly, SimCity is a very rich game environment and could attract lots of very smart designers to make it more extensible and malleable.
To get Alex started, we would need to articulate and list the "rules, heuristics, etc." that the classic SimCity uses. I don't know what form these are in, but I'll bet they look nicer in the reimplementations you've done than in the original game. You probably understand them better than anyone. In any case, if you are interested in this, Alex and I could come visit you (where are you?) and try to pick your brains about ways to give form to these rules that make sense in the world of relatively young children.
Cheers,
Alan
More great stuff from Alan Kay:
Hi Don --
I'm writing to applaud you for your plans to reimplement SimCity for children on the OLPC.
My main complaint about this game has always been the rigidity, and sometimes stupidity, of its assumptions (counter crime with more police stations) and the opaqueness of its mechanism (children can't find out what its actual assumptions are, see what they look like, or change them to try other systems dynamics).
So I have used SimCity as an example of an anti-ed environment despite all the awards it has won. It's kind of an air-guitar environment.
In the past, I tried to get Maxis to take the actual (great) educational possibilities more seriously, but to no avail.
Going to Python can help a few areas of this, but a better abstraction for the heart of Sim-City would be a way to show its rules/heuristics in a readable and writable form. Both of these could be stylized to put them in the child's own thinking and doing world. For example, just the simple route of making a drag and drop scripting interface for Etoys allows children to make very readable and writeable scripts and helps the children concentrate on what they are trying to do. A carefully designed object system (that is filtered fro children) can expose the environment so they can really think about it.
I'm not at all suggesting that Etoys be used here, but I am suggesting that some deep design be done to come up with a "behavior modification interface" that allows real creativity on the part of the children. So it is much more than stringing black boxes together or having to deal with fragile procedurals.
I sense that you have some interests in making SimCity really a microworld for children's learning and exploration from reading your webpage.
Children in 4th - 6th grade can do a lot here if they are given a good UI and tools. So, we could think of part of this project as a "pre-Python" UI.
Scalability and non-scalability of ideas are interesting. Rocky's Boots is still one of the best ever games that provide profound learning experiences. The extension of this to Robot Odyssey didn't work because the logic and wires programming didn't scale well enough -- the bang per effort dropped off precipitously. I was Chief Scientist at Atari at that time (Warren Robbinet worked for me) and I worked with TLC to try to get them to realize that something like Logo, or even better, a rule-based robot programming system, was needed. The failure of Robot Odyssey really pained me because I thought that the concept of this game was one of the best ever (still is). But it just needed a much better notion of how the children were going to program the robots. I think the same goes for SimCity.
Cheers,
Alan
I like the idea of forcing people to stay married, as long as you let gays get married and force them to stay that way too. Let's start with the current crop of Republican presidential candidates. Disqualify all the ones who've been divorced, and all the ones who are secretly gay, and you wouldn't have any Republican presidential candidates left!
-Don
You are totally talking out of your ass, and you're afraid to post that tripe under your own name, because you know you don't have a leg to stand on, and can't support your side of the argument with verifiable facts. Why don't you sign into Slashdot under your real name, and post some references to back up the ridiculous claims you're making? Just because you say it but won't sign your name to it, doesn't mean it's true.
-Don
First, the legal details:
The GPL source code version of SimCity will not be called "SimCity", but we will use the SimCity source code to make a city building game called "Micropolis", which was the original working title of SimCity. That's because EA reserves the right to review and QA the official version of the game that's published under the name "SimCity" on the OLPC. So we can make improvements to the TCL/Tk version of Micropolis (based on the GPL source code), and submit them to EA for review and QA, which if they approve, will be used as the officially branded version of SimCity for the OLPC. It will be the same code, but the only difference is the name, which EA understandably wants to protect, be ensuring the quality and integrity of OLPC SimCity.
That said, I strongly encourage you to improve the TCL/Tk version of Micropolis, please! I think it would be great to keep the TCL/Tk development line of the code alive and strong, and modernize and improve it! It will be a fair amount of work to upgrade it, but that will be a lot less work and time than recasting SimCity in Python will require. Of course both projects can go on in parallel, learning from each other. Since the Python version won't be ready for a long time, we can submit an upgraded version of TCL/Tk Micropolis as the official version of OLPC SimCity, based on the latest TCL/Tk, including further OLPC/Sugar integration \(like making sound work better, supporting the gamepad, etc).
TCL/Tk is a very cleanly written language and well designed user interface toolkit, which has grown and improved a lot since I took a snapshot of the code from around 1993 to use for SimCity. I have found memories of meeting Professor Ousterhout in his office on the UC Berkeley campus, and showing him the TCL/Tk version of SimCity, and thanking him for the great work he'd done and shared for free.
The reason I want to convert the TCL/Tk version of SimCity to Python is because it's the "official" programming language for OLPC Sugar applications. In the short term, the OLPC will include applications like SimCity that uses TCL/Tk, eToys that uses Smalltalk/Squeak, etc. But in the long term, new software developed for the OLPC should be written in Python, to save memory by sharing modules, and to use great modules like Cairo for scalable stencil/paint graphics, Pango for rendering internationalized text, and all the Sugar modules written in Python.
Despite not being written in Python, eToys and SimCity are included with the OLPC. The point is to educate and inspire people to develop new and better software, to bootstrap the software of the future. I'm compelled to move forward and develop a new visual programming language, better and easier to use than Python or TCL, that kids can use to reprogram the behavior and create extensions to SimCity, and even write their own games and who knows what else.
-Don
Alan Kay wrote this on the OLPC Sugar mailing list, about implementing visual programming languages like eToys in Python:
Guido knows that I've been advocating that the Python folks should do Etoys or a very Etoys like environment in Python (and that the rest of the OLPC be given an objectification and media and scripting integration that is Etoys like).
However, there are simply zillions of things left to be done everywhere for OLPC so the first round of SW on the XO will be more of a gathering of "suggestive" features and abilities (of which Etoys is just one). That seems fine to me.
Viewpoints Research (our little non-profit) doesn't have any "ego or identity" staked around whether the children's authoring environment is Python based or Squeak based. I have said many times that, if the general integrative base of XO is to be Python, then the Etoys-like authoring should be based in Python also.
However, I will personally fight to the death to make sure that there is a children's authoring environment that allows even young children to do simulation style programming with very ri
Well if you're not a Christian conservative, then do you support Bush and Cheney who pander to them? Do you believe they're only lying to the gullible Christians in order to get votes? Or do you think they believe what they say? In either case you're an idiot to support Bush and Cheney if you disagree with the religious views they claim, but you're not a Christian.
-Don
There is no original article. The onslaught of hyperlinks was the article! What's wrong with links? You don't HAVE to click on them. Or does your OCD torment you over every link left unclicked?
-Don
Hey, why don't you help Mitch Bradley port his excellent Forth system to run on the OLPC? It already does, but it's in the OpenFirmware boot ROMs, so it runs before Linux even boots. But the same Forth system runs quite nicely under Unix as well (without all the direct hardware access :), and is quite luxurous. I used it on the Sun 3 and Sparcstation years ago, and worked at Sun as his summer intern on CForth, another portable Forth system. Mitch is one of the best and most accomplished Forth programmers on or off the planet. If you want to do Forth on the OLPC, Mitch Bradley is definitely the dude to talk to about it!
-Don
Sure you're serious. And you probably take Stephen Colbert at face value. You conservatives are so funny, but totally out of touch with reality. You impeached Clinton for a blow job, but then you defend Bush no matter what he illegal and immoral things he does. And you all call yourselves Christians. Ha.
-Don
From: Designing User Interfaces to Simulation Games. A summary of Will Wright's talk to Terry Winnograd's User Interface Class at Stanford, in 1996. Written by Don Hopkins.
-Don