Oblomovka

2003-08-14»

Quinn told me two years ago that the basic problem with the global economy was that the idea of money was broken. Around the same time she announced that junk DNA had to have some sort of function, despite what current theories indicated. I humour Quinn on these and other theories, as long as she doesn't talk so loudly that the geneticists and economics professors at the next table hear her speak that way.

Today I discover that a) some scientists are coming to the same conclusion about junk DNA, and b) the guy who co-designed and implemented the convergence mechanism for the Euro, and co-founded one of the largest and most successful currency funds, Bernard Lietaer, agrees with her about money.

I can only conclude that those bastards were scribbling notes on their napkins all along.

Is it bad to secretly wish Schwarzenegger gets in, just because of the excellent rewrites of California Uber Alles it will encourage?

2003-08-13»

The co-location company that hosts a community server with which I'm involved went bust - at 11am yesterday. Without telling anyone, including their own tech support and hosting facility.

I suppose we could have predicted this. We've been trying to pay them for for a few months now, with very little success. Hard fast rule of e-commerce: if you make it impossible for people to pay you, and yet you are expecting to be paid, something bad will happen.

We can tell the precise point at which the co-loc company ceased to co or loc, because at that moment our machine vanished off the Net. A lot of heavyweight sysadmin types run their mail from this box, so turning off the packets is a bit like throwing up the BOFH-Signal into the sky. Heads turn.

First step when a hosting company spontaneously bankrupts: get your box out of there before the creditors mistakenly melt it down for slag. A fantastic friend who runs her company from the box ran cross-town to airlift it out. Serendipitously she bumped into a guy who is running a hosting outfit in the racks upstairs. He's an ex-employee of our co-loc, and realising what is going on, kindly takes us (and I'm guessing several other bedraggled servers) on board and plugs us into his network.

He's given us a few days leeway to sort ourselves out, but - always assuming this all hasn't been part of his evil masterplan - I think we'll go with him. He's still at that phase where he knows all his customers and answers the phone himself. I still have problems explaining to people why little ISPs like this seem to work better than big ones. I guess, if I wasn't so dog-tired, I'd say that the economies and diseconomies of Internet services are shaped like a big mexican hat. You can scale up pretty quickly, and then it all goes to shit until, if you're lucky, you sell out to someone big enough to run matters properly again. You can either be Henry Ford or William Morris, but you can't be Mr In-Between.

2003-08-11»

Well, the weekend's over, and I'm sticking with Python. I caught up to where I was with the Perl implementation pretty quickly, and in far fewer lines. To be fair, I think a sizeable part of that may have come from me using the Test::Unit / unittest suites to do my test-first development. Both are derived from Kent Beck's Smalltalk framework and are very OOP-oriented, which favours Python over Perl5. I was beginning to stumble into some hairy data structures with Perl, all of which turned out a lot simpler-looking in Python. That, and I'm doing some email mangling in this part of the coding, and Python's built-in email module has much of what I need. There are some even better libraries in CPAN (as usual), but I was greatly spoilt for choice and little spoilt for time.

Anyway, this still leaves the problem of a decent templating solution for Python. I got loads of very useful suggestions, which I'll attempt to summarise here:

Aaronsw wrote back to say that he uses cheetah and sqlobject . He wrote back suspiciously fast. I think it may have been one of his bots talking. Both cheetah and sqlobject look tempting - especially as sqlobject supports sqlite, and I'd like to stick with that very basic SQL implementation until my lusting and hunger for relational database power become known to more people, as the Boney M song has it ^[1].

Noted real-world Python user Zooko said that the Mnet gang would probably be going for whatever templating Twisted had. Then he confessed that they probably wouldn't have much need for templating at all. His most abiding point was, I think, to exclaim w00t python w00t, which I took to heart.

Jonathan Moore says "all the cool kids have quit templating and gone to XSLT", which I subconsciously knew. But then if the cool kids jumped off the Post Office Tower, would I join them? Naturally, because cool people are usually into BASE jumping and stuff like that, and would be cool enough to help me with basic safety precautions. And, besides, how cool does jumping off the ... I digress.

My thoughts: XSLT is one of those technologies best kept for either times when someone is paying you to learn it, or you've lied on your resume and said that you wrote the W3C spec, and now you find you have the job of your "dreams". I'm in neither position, for once.

David Jeske wrote an informed recommendation for ClearSilver, which isn't surprising as he's been hacking on it for yonks. ClearSilver looks really interesting. It has a nice, simple, theory behind it. The original, closed-source version drives Yahoo Groups . Jeske and friends' open source re-implementation is used at Weather Underground. It doesn't get much more hardcore than that.

Van Gale backed up the Clearsilver recommendation, and also put it in a plug for Cheetah and TAL (the Zope templating system untimely ripped from the rest). Van has a soft spot for PyMeld too, but wasn't sure if it could cope with high-performance sites. Richard Jones likes SimpleTAL.

When I hit my templating problems, I think I'll hit them back with either ClearSilver or Cheetah. Which one will depend on the size of my problems and on what speed I'm plummetting into disaster at the time.

2003-08-09»

I've got a new medium-sized project to be working on. It's just me coding it at the moment but at some point (if it comes together), I do want to be able to pass it on to someone else, preferably without apologising first.

I'm utterly torn between Perl and Python. My first choice in this case would be Python, because bad Python code doesn't seem to be quite so personal. I've seen people spit blood at other coder's Perl, just because it's not the way that they would do it. Perl demands rather more sympathy with your predecessor than does Python. With Python, it's just more code to stare at.

That said, your successor does need to actually know the language. Most of the people I can imagine maintaining this code will know Perl but not Python. Python doesn't take that long to learn, but reading Python to take on someone else'se project just isn't much *fun*. Sitting down to learn someone's Perl, while tough, does teach you about the way they were thinking when they wrote the application. Python's clarity, I think, cuts down on its expressiveness in depicting why certain decisions were made. When I had to hunker down and learn POE or Moveable Type, for instance, I came away with a very deep understanding of how it was supposed to work. It was fun, albeit time-consuming. I sometimes have problems doing the same with slabs of Python code, just because they can be very lacking in personality.

That said, I'm not paid to be a programmer. What is fun is a hobby can be skull-crackingly frustrating in a job with a deadline.

I eventually made the decision to go with Perl and Mason - mainly because of Mason. I know a lot of people who know Mason who I can ask when I got stuck, and there's a fair bit of this code that will end up being on a Website, eventually. There don't seem to be any Python templating solutions that stand head-and-shoulders above the crowd (and whose implementors I know). I felt that going with Perl would provide my successor with a clearer understanding of the whole project, and not necessarily lock me into an immature Web tech that no-one knows.

Now, a couple of days into it, I've begun to seriously reconsider. I'm nowhere near the Mason bit of the application, and I'm getting continually bogged down in Perl style issues that really don't have anything to do with what I'm trying to write.

To be honest, I think this is my Perl rustiness kicking in; and I think it may go away after a few more days hacking. Worse, though, is the effect of something I thought would be a real boon - CPAN. There's a bunch of useful utilities there that I'd love to suck in and use in my program. But they all have different idioms - all of which I have to sit down and learn. Plus there's the whole dependency issue: sooner or later I'm going to have to install all of this on the working server, and there's a real penalty to be paid for being dependent on a lot of scattered Perl modules. Will they work? Will they still be maintained? Which of alternative implementations should I choose?

So, I'm going to spend this weekend having a bash at a Python version. My XPish working style has recently become dependent on a lot of Python features, and the Python core library has most of what I need in the way of modules. I really haven't written much code, so there's not much time lost. And I figure that if it all goes to hell in Python too, I'll learn my lesson quickly and go slinking back to Perl with my tail between my legs. Plus, no-one's paying me to be decisive here. As if anyone would.

I still don't know of a good, solid, Python templating language though. Or rather, I know of three or four, but I don't know anyone who has used any of them seriously in high-volume production environments. I'm not interested in Zope or Twisted (tempting though they are) because I'd be obliged to wrap my solution in their terms, but I really don't know whether I should go for SkunkWeb or WebWare or Spyce or any of the others.

Right now, I'm delaying the decision until I'm a bit deeper into the code. Which feels a bit uncomfortable. But under the terms of XP, perhaps forgiveable.

2003-08-07»

There's a US federal law which lets individuals prohibit companies from sending unsolicited sexually provocative or erotically arousing mail. According to the postal ruling implementing this law, ""Postmasters may not refuse to accept a Form 1500 because the advertisement in question does not appear to be sexually oriented. Only the addressee may make that determination."

Kayne McGladrey is doing the obvious thing: declaring that all those Sharper Image catalogues and coupon collections are quite filthy, and should be exorcised from her letter box. Instructions included.

Learn something new every day, with the vim.org tips RSS feed. (Thanks, Phil!)