Where Everybody's Crazy

Since I had all this time on my hands, I thought I'd have a quick look at how Parrot has been getting on in the six years of my absence.

There's been a series of articles on Perlbuzz and other places about how Perl 6 is nearly there, how Perl 6 is already there, how Perl 6 is dead in the water, and so I thought I'd go see for myself. My old SVN login still works; thankfully it looks like none of my original code for Parrot remains though. :)

I've come away thinking that Parrot is hugely important. I also think things have come a heck of a lot further than anyone outside the project could imagine. There's a perl6 make target which builds Perl 6 binary, for heaven's sake, and it runs real live Perl 6 code. It's there, right now. But that's not what excites me.

What excites me is the Parrot Compiler Toolkit.

Around the time I stopped working on Parrot, people were saying that Perl 6 was going to be written in Perl 6, and I was, well, let's say, a little sceptical. But you know what? They've done it. I am astounded. It is lovely.

This afternoon Will Coleda checked in a simple stub implementation of LOLCODE. The implementation consists of three files. The first is the grammar. This is written in Real Life Perl 6 Rules. That was pretty impressive. Then there's the actions file, which builds the Abstract Syntax Tree. This is written in a language called "Not Quite Perl", which is as close to Perl 6 as makes no odds. Finally there's the implementation of the builtins, which is written in Parrot's intermediate language. This is where you implement any custom runtime operations that Parrot doesn't already know about.

I spent two hours this afternoon basically starting from zero knowledge of the Parrot Compiler Toolkit and added a bunch of features to Will's stub, including variable assignment and access. That is astonishing.

The bottom line? The takeaway? It's really, really true. Parrot lets you implement your own languages using Perl 6 rules for the grammar and Perl 6 for the compiler.

Today it is so cold that I am doing all I can to stop myself from declaring the day over, and curling up in bed with the fan heater and the electric blanket and a half-dozen other blankets on top. And it's only six pm. This is not good.

Yesterday, though, it was warm enough for me to have a Hacking Evening. One of the things I needed to do here was install Request Tracker for sorting through our short-term worker approval process. (I'm sure there will be other uses for it, but it's usually one obvious use that lets you shoehorn it into an organisation and gets people to realise how useful it is. Heh, heh...)

Anyway, since installing RT has gone from being the full-day brainache-inducing nightmare it used to be, to now being a matter of a couple of commands - which takes so much of the fun, and certainly most of the consultancy fees, out of it - I had to do something to complicate things again.

I got scared about the fact that one Apache server has PHP, Perl, SSL and a host of other extensions hanging off it, and particularly how one of my users' experiments with mod_perl has been known to take the whole server down, and decided to try something a bit more flash.

This blog and the web sites hanging off its server are actually front-ended by a Squid reverse-proxy to speed things up a bit and take some of the pressure off serving mostly-static files. I thought I'd do something like that but with not just one but several back-end web servers; where I used to work we use something called Squirm to handle redirecting a bunch of virtual hosts off onto their own servers. (That's due to using Zope, which runs its own web server for each application you set up. If you want more than one application on the same IP address, you have to find your own way to multiplex them.)

Anyway, I was just starting moving vhosts onto their own server instances, and working out nice ways to do the Apache configuration and init.d scripts and everything, and then I realised that squid and squirm is big heavy overkill anyway, and I should check out this thing called Perlbal that I'd heard various folks raving about.

At that point I realised that I didn't even need a separate server for all of the static-content web sites - Perlbal can serve them itself using sendfile goodness, which makes it extremely, extremely quickly. RT has its own standalone web server, a bit like Zope, so I got Perlbal to proxy that as well; the way I see it, the less Apache, the happier my servers are. And everything else, all the horrible PHP stuff, gets proxied back to Apache normally.

And like I said, it's extremely, extremely quick. Both thumbs up.

Looks like I found a perfect approach to combatting comment spam...

Someone back at WEC UK asked for my views on Internet evangelism. I wrote back a long response which was less than favourable, so pretty much killed the conversation, but here it is for posterity...

See, I'm not sold on Internet evangelism because, as we know, the web is largely (although not exclusively) a pull medium and evangelism is a push medium, so I just don't believe they can fit well together. In other words, in evangelism you want to be going out to reach people, but in constructing web sites you're waiting for people to come to you. Developing material is fantastic but bringing people into contact with it is the tricky part.

Yes, I've read the "bridge strategy" stuff but all the strategies there still feel completely churchy to me. I honestly don't believe that people search the web to find articles about the meaning of life. "Who is Jesus Really?" is, I'm sure, a brilliant site, for people who are already interested in the answer to that question. But that's not the question that most Japanese are starting from. (This can also be understood as a criticism of Alpha in Japan.) Looking over the search keywords people use to stumble on my site in Japanese backs me up here; there are one or two people looking for things like "how to love myself" (self-image is a big issue here) but the vast majority are obviously already Christian, searching for Biblical characters and so on. So if you're going to do this, I reckon that the emphasis really needs to be on constructing communities that people will find and join, or more likely presenting apologetics in existing forums, which is what I tend to do a bit in Usenet, IRC and the like...

I've never seen a satisfactory answer to the question "how are you going to build your audience?" in the web evangelism field. I have seen people throw around terms ("social networking!") as if they were magic cure-alls, instead of means to connect people in the usual way that people connect to people. It's got to come down to going out into the highways and byways, and that takes effort and building personal relationships, the usual missionary grunt work; simply throwing content at the web and hoping it finds its intended audience isn't going to cut it for me.

So at the moment I'm more interested in mobilizing the Internet for serving the existing church community than primarily for evangelism. Social networking for accountability and prayer partnerships, that kind of thing. Mitsuo Fukuda is doing this with his businessmen's groups via SMS; Masaru Aoki of the Diaspora Network Japan runs online Bible studies for Japanese overseas and returnee Christians, and follows them up with Skype discussion and prayer. This is more the kind of thing which I like, since you already have a natural community that you're serving.

This isn't to say don't do it. That's not what I mean. We should try absolutely everything, and maybe save some. I'm sure it does save some. Just saying why it doesn't turn me on much.

I've been feeling like a compete sloth this week and not getting very much done, but today I managed to do a great deal of work, and am happy to announce the 1.0 release of Songbee. There's an awful lot changed since the 0.1 alpha, including:

• Search bar during worship
• Song import and export
• "N" key to go to the next logical part of a song
• Projector display preferences to customize how the worship screen looks
• Full-screen mode

The only down side is that as yet the manual hasn't caught up with all this new stuff, nor has the Japanese translation of the user interface.

I'm having some problems with the Windows build, due to this bug in xulrunner. I don't know a workaround for it at the moment, but if you have one, please share it with me. OSX and Unix versions work fine.

As usual, bugs, feature requests, etc., let me know.

Update: I think I've fixed the Windows issues. If someone could give it a download and have a try, let me know. Thanks!

So I launched a new site today to hold all my free literary output. It's not so much a question of making it all freely available; rather it's an experimental new writing environment for me. Normally when I write stuff it's either in TeX or DocBook XML, and despite trying all kinds of folding editors and what have you, I still get a bit lost in the structure; with Mediawiki I should be able to pick out sections that I want to edit in isolation, but still be able to see the big picture altogether.

There's a few bits of good Perly stuff in there, but I'm currently working on one book, a translation of Mitsuo Fukuda's book "Developing a Contextualized Church as a Bridge to Christianity in Japan". I'm reading through this every week (although not so much this month) with my pastor anyway, so translating it forces me to actually sit down and understand it before I go into my Friday morning meetings!

I set myself a list of goals for today - I often do this, as it's one of the few ways I can ensure that I feel satisfied with my day. Even if it is, theoretically, a day off. Protestant work ethic, etc.

Today's goals were:

• Customize WEC's Memories installation to do everything that they expect before going live
• Make Songbee import song library files
• Write the structure to next week's sermon.
• Upload photos from yesterday's day in Kobe.

I did them all, which is half way satisfying, and half way makes me think "So what should I do tonight, then?" Maybe I need to learn how to relax... :)

So, like I said, I have this week theoretically off. It's given me a chance to catch up on some long-standing programming todos.

I've found a great new place to work; I was cycling around on Monday trying to find a Starbucks in town or some other laptop-friendly cafe, since it's way too hot to be working at home. Eventually I remembered about our local Internet cafe, which looked very cool from the outside, and I had been meaning to investigate it. It promised a lot: open 24 hours, showers, karaoke, darts, pool, PS2s with plenty of games, massage chairs, free drinks, a comic library, wide-screen TVs, and a lot more.

It's all true. Plus free ice cream and miso soup! And it's cheap. But all the gimmicks in the world aren't what I'm paying for: I'm paying for a place to get out of the heat and concentrate and get some work done. And it's really, really good for that.

So what have I been up to? I've been making some major changes to Songbee, including finally getting around to providing customization preferences for the display, and starting to sort out song library imports. This has seen me digging into Javascript and XPCOM in more and more complicated ways, and I'm actually beginning to feel comfortable with it now. There are still some annoyances, but they're more architectural things than failings - I still don't get on with the fact that a lot of the working is asynchronous; if I add a stylesheet into a document, my code continues running even though the stylesheet hasn't finished loading yet, and there's no easy way to wait for it. I've hacked around it, but that kind of thing. The preferences thing is a lovely bit of JS/CSS/DOM manipulation. Go look at it.

I've also been hard at work setting up a photo sharing site for WEC Media, and making some changes that they want. Aside from programming, I've been putting together a flyer for a leadership seminar I'm giving at the end of September.

I have had time off too, even though I still feel guilty about taking some. I took all yesterday to play go, and tomorrow I'm off to visit Kobe. Then it's back to work again, with yet another camp...

So today I've been doing work on Songbee, amongst other things; one of the areas I've been looking at is whether or not the Japanese localization of it works properly. It certainly doesn't on Mac OS X. Even if you select Japanese as your preferred language in Mac's System Preferences, the Songbee interface comes up in English. This is counter-intuitive, and means it doesn't work like any other OS X application.

Now it turns out that this is a problem with the Mozilla platform. It doesn't ask the operating system what language the user wants. (Although this isn't hard to do; here's some code which does it.)

It also turns out that this problem has been known about since 2001. I don't know, maybe it's just me, but I think if you have important bugs (of course it's important, because it's important to me dammit! ;) sitting around open for six freakin' years and nobody's touching them, something's gone wrong with your bug tracking system. Part of the problem, I guess, is that Mozilla gets a lot of bug reports, and there's enough work to do with triage and fighting fires trying to keep up rather than go look at old stuff. So this is no criticism of the Mozilla project - perhaps the only way old bugs will be looked over will be if someone volunteers for the boring and thankless job of trawling through a very large and very old bug list and working out which ones are still important. But volunteers for boring and thankless jobs are, for some reason, really hard to find...

A few weeks back Tim wrote about mobile phone novels. On Monday, I got myself a mobile phone - not a particularly good one, but it's cheap - and what did it come with but a Java comic reading application.

Mobile phones are a great medium for comics; as was pointed out on Tim's article, lots of people spend lots of time on trains, often reading comics or messing with their mobile phones - why not combine the two activities?

Additionally there are a couple of features of Japanese comics which make an electronic reader work very well. Scott McCloud's classic Understanding Comics demonstrates that Japanese comics use two effects a lot more than Western comics: bleeds - that is, panels which escape the gutter and bleed out to the edges of a page, to give a wider impression of a scene, and aspect-to-aspect transitions - several panels providing a range of views across a scene. So, how do you do a bleed on a tiny mobile phone screen? You get the same effect as a bleed, and as a aspect-to-aspect transition, by side-scrolling the panel across the screen. Side-scrolling also helps the reader to follow the order of dialogue when there is more than one speech-bubble in a panel.

And of course there are other tricks you can do with a mobile phone comic. I had just got to the bit in the story when a gunshot rang out, and at the same time, the phone vibrated like hell...