WEBVTT 00:00.000 --> 00:04.780 You can check out on video of all the Japanese people . 00:30.000 --> 00:52.000 Good afternoon fellow dinosaurs. I would like to start with the disclaimer. I'm a nitriptist. 00:52.000 --> 01:01.000 I'm a nitriptist. I would like to start with the disclaimer. I'm a nitriptist. 01:01.000 --> 01:06.000 I'm not a nitriptist employee. It's important. 01:06.000 --> 01:12.000 This presentation is mostly recycled material from Neo4f 24. 01:12.000 --> 01:18.000 I would not assume that you guys are familiar with the fourth programming language though. 01:18.000 --> 01:25.000 What's interesting here is the development methodology and the scope of this endeavour of course. 01:25.000 --> 01:31.000 The development never is a straight arrow from nothing to whatever your goal is. 01:31.000 --> 01:39.000 It's more like a revolutionary process, like most of you know probably. 01:39.000 --> 01:47.000 The target, the target is this little board which I designed. 01:47.000 --> 01:58.000 This old started at the end of 2018. After having experimented with the Arduino AVR platform for one, 01:58.000 --> 02:05.000 the fact needs to go back to my digital electronic routes because that's my initial training. 02:05.000 --> 02:13.000 I decided to put together my own 6809 single board computer which is what I just showed you. 02:13.000 --> 02:25.000 I also had a 35 year old printout of a fourth implementation I wrote back in 1983 for the Z80. 02:26.000 --> 02:33.000 Eventually the combination of the two gave birth to a platform which I call Z79 fourth. 02:33.000 --> 02:41.000 This is a sort of some sort of a pun because Z80 was the original target. 02:41.000 --> 02:50.000 And the 79 is the original fourth standard specification I tried to implement. 02:50.000 --> 02:58.000 So this is a minimalist platform. It has the basic functional literary credits for running fourth. 02:58.000 --> 03:03.000 Mostly from rum. Some of it comes from compact flash. 03:03.000 --> 03:10.000 It's USB powered which is convenient. You can use any phone charger to power it. 03:10.000 --> 03:17.000 And the serial console is over an asynchronous serial line. 03:17.000 --> 03:29.000 For this exercise I used the ANS94 firmware variant which is kind of a modern standard. 03:29.000 --> 03:38.000 And most importantly it's compatible with GN4 which allows of course for cross-platform development. 03:38.000 --> 03:45.000 So I can develop onto Linux and GN4 and run on this little guy. 03:45.000 --> 03:52.000 And obviously the Linux development is much more comfortable. 03:52.000 --> 04:00.000 The target is the target terminal is the deck VT4 20. 04:01.000 --> 04:06.000 So the reason for that is that USB is nice. 04:06.000 --> 04:14.000 In my opinion if you want reliable flow control you want RS 232. 04:14.000 --> 04:25.000 So instead of using a 500 watt PC as a console emulator I bought a deck VT4 20. 04:25.000 --> 04:29.000 Why emulate when you can have the real thing? 04:30.000 --> 04:35.000 Now the VT4 20 is a monochrome text terminal. 04:35.000 --> 04:39.000 And Pacman is a graphical colleague. 04:39.000 --> 04:44.000 Somehow, which is a typical built. 04:44.000 --> 04:56.000 Fortunately the 420 supports a function called deck download which allows for unit defined fonts. 04:56.000 --> 05:04.000 So this was an opportunity for custom sprites which I took advantage of. 05:04.000 --> 05:12.000 I decided to implement differentiated aspects for each ghost as a substitute for colors. 05:12.000 --> 05:24.000 Because you don't have colors and you want still to be able to tell which ghost is which because they have different behaviors. 05:25.000 --> 05:32.000 The basics of Pacman everybody knows Pacman in his audience as opposed to what he wouldn't be here. 05:32.000 --> 05:46.000 Namco is the name of a Japanese company which produced Pacman as an arcade console game back in 1979, the good old years. 05:46.000 --> 05:50.000 I'm sure you'll know how this is played. 05:50.000 --> 05:53.000 There is a amazing venue. 05:53.000 --> 06:03.000 Interspersed in the maze are collectible items which are either fields or power pellets. 06:03.000 --> 06:13.000 The power pellets of the virtue of giving Pacman some immunity against the ghosts for some limited period of time. 06:13.000 --> 06:21.000 But they also carry a higher point of value. 06:21.000 --> 06:35.000 So ghosts are trying to get to Pacman before he can complete going through all the collectibles in the maze. 06:36.000 --> 06:46.000 The original mainstapology is some sort of a singularity here and here you have a tunnel. 06:46.000 --> 07:04.000 So I decided to stay away from this singularity and I went through a different topology which is the standard Linux version by RORFORNICE. 07:04.000 --> 07:08.000 This is here, X11 Polyx, C++. 07:08.000 --> 07:18.000 Also, Cigwin Windows, the concession of the times are supposed. 07:18.000 --> 07:28.000 So I selected this topology and also the object oriented approach which was really, really useful. 07:28.000 --> 07:32.000 So method, yeah, you need one. 07:32.000 --> 07:40.000 As Woody Allen once famously said, confidence is what you have before you understand problem. 07:40.000 --> 07:50.000 Woody Allen obviously was an engineer, just not the technical kind, the human relationship kind. 07:51.000 --> 08:07.000 My approach was based to greatest extent to on Jamie Pittman's the Pacman dossier, which is a comprehensive reverse engineering document about the original Z18 Namco implementation. 08:07.000 --> 08:10.000 So there are two levels here. 08:10.000 --> 08:15.000 Presentation layer which is the visual aspect of things. 08:15.000 --> 08:22.000 This entitled, obviously, the design of a user defined font. 08:22.000 --> 08:35.000 That took me 12 days full time, so it's significant and it's about 30% of the code if you omit the comments. 08:35.000 --> 09:00.000 The business layer is the game logic and we can further divide that into basic business and AI integration, which means that each ghost has a different targeting behavior when it is in chase mode. 09:00.000 --> 09:15.000 Eventually I've ported this fourth code to see because I wanted to run under open salaris and open VMS as well, and I'll get back to that later. 09:15.000 --> 09:22.000 So presentation layer, the basis. 09:22.000 --> 09:38.000 The deck VT420 offers a noticeable space of 80 columns and 24 lines plus 25 status line, which is not really addressable. 09:38.000 --> 09:51.000 If I elected to map its sprite to the double character, or horizontally, this required 30. 09:51.000 --> 09:56.000 Yeah, I turned myself and I should be all right. 09:56.000 --> 10:12.000 So, this required 30 double with custom character fonts to be designed and as a result, you end up on the x-axis, its physical column is half a cell. 10:12.000 --> 10:25.000 And vertically, a virtual row is a triceterminal row, so every club cycle you move in the virtual space one step, but physically not. 10:25.000 --> 10:32.000 On the next club cycle you will move physically one row up or down. 10:32.000 --> 10:42.000 So, in actuality entities are moving in space that is four times bigger than the game space. 10:43.000 --> 10:48.000 Now, font design. 10:48.000 --> 11:03.000 On the 420, a character as a food cell, as the deck documentation state, is defined as a 10 pixel by 16 rows. 11:03.000 --> 11:18.000 So, you have to for each element of the maze and of the moving entities, moving objects, which I call entities, you have to come up with a design like this, and there are 30 of them. 11:18.000 --> 11:21.000 This is a four-scode. 11:21.000 --> 11:26.000 It's partly interpreted, this is all interpreted. 11:26.000 --> 11:37.000 For those of you who know basic, it's like data stuff, and this is what gets sent to the terminal to define this double character. 11:37.000 --> 11:46.000 So, it's defining three groups, which I call six holes, the significant bit is at the top. 11:46.000 --> 12:03.000 So, if you define your character like this, you beat groupsy, simple enough, but time consuming, as well, and not really exciting work. 12:04.000 --> 12:09.000 What should be on your own? 12:09.000 --> 12:12.000 All right. 12:12.000 --> 12:14.000 Objectary-oriented design. 12:14.000 --> 12:17.000 Yeah, it's fairly simple. 12:17.000 --> 12:24.000 This is centralized, a state information, essentially. 12:24.000 --> 12:35.000 It uses 4th 2012 structures with pointers to methods. 12:35.000 --> 12:37.000 Yeah, this is very simple. 12:37.000 --> 12:39.000 There's no inheritance. 12:39.000 --> 12:41.000 This is the method in vocator. 12:41.000 --> 12:44.000 Very simple stuff. 12:44.000 --> 12:52.000 I like to keep it simple. 12:52.000 --> 12:55.000 So, yeah, all objects are equal. 12:55.000 --> 13:00.000 But Pac-Man is always processed first in the domain engine loop. 13:00.000 --> 13:08.000 Yeah, of course, it's instance zero by convention. 13:08.000 --> 13:13.000 Which brings us to scheduling. 13:13.000 --> 13:18.000 This stuff doesn't use interrupts or anything, it's just busy waiting. 13:18.000 --> 13:21.000 So, the main loop is like this. 13:21.000 --> 13:26.000 If we enter a level, we do level entry in it. 13:26.000 --> 13:29.000 Otherwise, we do regular scheduling. 13:29.000 --> 13:36.000 All the objects are invoked for the strategy method. 13:36.000 --> 13:43.000 And then we wait for a fixed period of time, time-sliced, clock period. 13:43.000 --> 13:53.000 The strategy method computes the next coordinates of each object in virtual space and updates it on the screen. 13:53.000 --> 13:56.000 So, this is basic scheduling. 13:56.000 --> 14:04.000 This works really well on a single task system. 14:04.000 --> 14:12.000 Yes, you can do real time when this baby and it works super well. 14:12.000 --> 14:13.000 But there's more. 14:13.000 --> 14:20.000 There's more because the reference documents, the Pac-Man dos, he says. 14:20.000 --> 14:27.000 At any given time, ghosts can be in any of three modes, which are scatter. 14:27.000 --> 14:30.000 In scatter mode, ghosts converge to their home corner. 14:30.000 --> 14:35.000 So, those are the full corners of the maze. 14:35.000 --> 14:38.000 Is it here? 14:38.000 --> 14:45.000 In Chase mode, it goes as its own targeting policy. 14:45.000 --> 14:48.000 Yeah. 14:48.000 --> 14:53.000 What should be all right? 14:53.000 --> 14:59.000 Which depends primarily on the Pac-Man's location, of course. 14:59.000 --> 15:07.000 I developed a time-based state automaton to implement this scheduling, which is specified in Pac-Man dos. 15:07.000 --> 15:14.000 Again, so a sub-scalular. 15:14.000 --> 15:23.000 Then the AI part, I don't like to change the term AI because it's not reading AI. 15:23.000 --> 15:28.000 So, it's ghost instance, as its own targeting policy. 15:28.000 --> 15:34.000 There's a field in the entity data structure that defines the instance number. 15:34.000 --> 15:45.000 And based on that, we select a target and we invoke a navigation routine that targets this specific target. 15:45.000 --> 15:49.000 So, again, nothing fancy. 15:49.000 --> 15:57.000 In fourth, we tend to do simple things efficiently. 15:57.000 --> 16:07.000 Now, the C port, yeah, I targeted Linux first because it's seem to be the easiest one. 16:07.000 --> 16:12.000 But actually, I've got easier success with Solaris. 16:12.000 --> 16:15.000 So, yeah. 16:15.000 --> 16:20.000 If you care about portability, you care about politics, but it's expensive. 16:20.000 --> 16:28.000 If you want to have access to the standard document, you have to pay 700 bucks. 16:28.000 --> 16:29.000 Yeah. 16:29.000 --> 16:34.000 Or I could use my money in various ways, I suppose. 16:34.000 --> 16:44.000 The AI typically advocates open standards, which means that if you want the real thing, you will have to pay some serious money. 16:45.000 --> 16:52.000 Terminal control, you need that because by default, a terminal will echo anything. 16:52.000 --> 17:00.000 So, you need to turn it off and switch to row mode. 17:00.000 --> 17:10.000 That was the initial design, then I switched to curses because that's the only way that the only thing that the MS can do. 17:10.000 --> 17:11.000 Yeah. 17:11.000 --> 17:18.000 A few primitive off-worth keywords, the system depended. 17:18.000 --> 17:22.000 And this is where most of the porting effort was. 17:22.000 --> 17:24.000 The MS, yeah. 17:24.000 --> 17:28.000 You have to use poll and curses. 17:28.000 --> 17:34.000 So, as a result of all this, you get three time response. 17:34.000 --> 17:39.000 Really good stuff under my board and Solaris. 17:39.000 --> 17:47.000 The next is not so great, you have to wait a few seconds before input is taken into account. 17:47.000 --> 17:57.000 And the MS is a big disappointment because there's a bug in the sea runtime library that prevents me from handling input. 17:57.000 --> 18:08.000 So, I can upload the phone display the maze, but the first time the program queries the keyboard status, it crashes with the EIO. 18:08.000 --> 18:20.000 That is bad, but I guess I'll have to wait a month or two before the VSI fixes the problem. 18:20.000 --> 18:31.000 So, this was possible thanks to Jamie Peatman, Patman Dosier, VT100.net, a Bible. 18:31.000 --> 18:43.000 And of course, the good people at Teshnisha University's dean for the new forth conclusions. 18:43.000 --> 18:46.000 Fourth, it's really fantastic. 18:46.000 --> 18:51.000 You define your own concepts and you are the limit of the program. 18:51.000 --> 18:58.000 Specification requirements? Yes. They're useful if they're done for you, like it was the case here. 18:59.000 --> 19:02.000 You can benefit from that. 19:02.000 --> 19:07.000 Standard, I can't emphasize enough for the value of standards. 19:07.000 --> 19:12.000 Standards means portability, interoperability, interoperability. 19:12.000 --> 19:17.000 And this is a great, of course, free standards are better. 19:17.000 --> 19:26.000 And I think a really great developer develops his own tools to get to where E or she wants to be. 19:26.000 --> 19:29.000 And that's it.