Oblomovka

2003-07-13»

I swear I had four ideas for mini-programming projects at some point last week. Now I find they've evaporated down to one. And even that one has already been thought up before by Kragen. But no-one, to my knowledge, has implemented it, and it's still a great idea.

I often find myself - as many did at OSCON last week - joining a network with a perfectly fine Net gateway but dead or unresponsive DHCP server. It'd be really useful under those situations to have a program that can deduce the gateway, network mask, spare IPs not in use on the Network, and optionally nearby DNS servers, Web proxies and authorised SMTP relays. You can work out much of these from judicious sniffing of traffic on the network; others you can deduce with some simple heuristics.

Kragen's thought about this a lot, but makes it sound a lot sneakier than it has to be, because he's trying to work out tricks to use in case you're plugged into a switch. On a switch, you don't see much other traffic. I'd be happy with the autoconfigurator if it only worked on hub-based networks (or wireless networks), where you can see most traffic already.

About the most potentially antisocial act this program would have to do is guess an IP address. If it chose one that was already in use, you'd disrupt that computer's connection. But a combination of sending out ARP requests, sending out test traffic, and being conservative in your choices (aiming for a reasonable distance away from present IPs, avoiding either edge of whatever netmask you've estimated the network to use).

2003-07-12»

Now I'm really peeved. I had to miss Miguel's keynote to write NTK. I spent the rest of the day regretting it, because Quinn told me it was very funny, and I like funny. But now I'm going to spend the next couple of weeks regretting it, because apparently they demo'ed dashboard.

A couple of weeks ago, when Nat Friedman first started hacking seriously on it, he wrote "I think this project has a lot of dorky-blogger-meme-virulence potential". I think it's more than that. This looks like the sort of app my friends would happily murder me for.

In a nutshell, it's a system-wide Remembrance Agent : a system that takes little clues from all the apps you're using and throws up suggested links to other items. So if I'm talking on IRC to, say, mouthbeef, my IRC client will send a cluepacket to dashboard, announcing that. Dashboard then echoes the cluepacket around to other listening apps, and they'll all chip in with their suggestions. "Mouthbeef", says my address book: I'm sure that's Cory's nym. "Cory?", says my mail archiver, "okay, here's the last five mails from Cory". "Right, I know that Cory writes for BoingBoing", says my aggregator, "so here's the RSS feed for that".

I've seen work like this from Microsoft Research, but it's always struck me as a bit too clever-clever. They always go for the angle of "well, you're makiing a lot of noise and not typing, and the phone is off the hook, so there must be someone else in the room, so I won't display these private messags"-type hints. It's far too deductive: it's smart for smart's sake. And their set-up always makes what I think is a key mistake, which is to work too hard to delegate decisions to the computer. I see this a lot in agent tech; all that bullshit about "My preferences show that you like tasty burgers, and we're in the neighbourhood of a Big Kahuna Burger joint. Let me show you the route.". No. I may well like burgers, but you'll show me the solution when I ask for it. Don't act like a personal assistant, making decisions for me. You're no good at that. Give me more information to make my own decisions. Increase my power, don't bleed it away.

Also the Microsoft stuff continues to have its head stuck right up the ass of corporate America. One of my big bones with MS stuff is that it always makes me feel like I'm eating out of the trash bins outside a cubicle farm. All of their software is designed to help busy executives plan their lives. Everyone I know uses it to try and write birthday cards and chat with their friends. When people use Microsoft Office they use it anywhere but in an office. Microsoft knows this - but it also knows that the money comes from their corporate clients, so there's a limit to how much it can bend its software toward a wider customer base. Ultimately when you use MS software, you're not the end user MS perceives at all: we're just living off the scraps Microsoft leaves out after feeding its big customers. This is especially true of their super-smart agent tech. Every demo I've seen presumes so much about how it's going to be used in an office environment that I can't imagine using it anywhere else. Actually, I can't even imagine using it in a non-WASP, non-North American office a lot of the time. I'm sure they'll try and fix this, but their hearts won't be in it. WASP America is their heartland.

Anyway, back to Dashboard. I like the look of Dashboard partly because it feels very informative, rather than anticipatory. It's not really making deductions about what you want to do, but throwing you extra information about what you're doing now. Think status bars versus paper clips. Also, the very fact that it's being hacked upon by the GNOME folk means that it's already working in an environment much closer to my own.

The first of those is, I think, a permanent advantage. Ever since I saw the Remembrance Agent , I've wanted something like Dashboard. It just seems to be the right idea for me, and I think for others too. It's what I liked about Lotus Agenda ; it's what I anticipate liking about Chandler.

The second is selfish. Just because Nat and Miguel have usage patterns closer to me than to a CEO isn't a universal good. The free software's bias towards hackers is no better than Microsoft's bias towards companies. But I do think there's more of a chance of software like Dashboard gravitating towards other lifestyles. Not really because I think that the free software movement is particularly good at accepting different user behaviours, but because Microsoft is so spectacularly bad at it that there's a real vacuum in the market. It's the Achilles heel that Apple is so carefully exploiting these days. I think that the open source movement would have to positively work at being close-minded not to win here.

There now; I've written a huge amount on a piece of software that I haven't even downloaded and run yet. I get the impression it's not really in a workable state yet, and anyway I'll have to get Mono back up and running before I can play around. And I'll probably have to write a set of clue-generating backends for vim or something. I'll do it in my copious spare time and let you know.

Okay, I mustn't gawp, and it's hardly germane to what she's writing about, but I still think it disarming to discover a Sixties Oxbridge, Hitchens and Clinton reminiscing, New Statesman crowd posting on Livejournal. And they have a little discussion afterwards! And they all have pre-raphaelite icons!

I can't quite put my finger on why this is so pleasantly unsettling: it's like seeing two class systems mix it up. It's like some online equivalent of Jeffrey Bernard's Low Life column, but you get to hear the discussion in the pub afterwards. I like it, but then I like most things that give me a headache afterwards.

Of course, learning more about Roz Kaveney, I'm no longer surprised. Anointed to be a gatekeeper.

2003-07-11»

The lightning talks worked very well at OSCON (and proved hilariously stereotypical : the Python talks were well-ordered to a Netherlandish extent, the Emerging Tech ones were largely performed by people with brightly died hair, and the Perl talks were even more ADHD than you'd imagine). At the end-of-conference press overview, Nat said he was going to go for lightning keynotes next year: 800 pundits in half-an-hour.

The five-minute talk that surprised me the most (apart from the Chinese rap) was Blog Census. Maciej Ceglowski has been writing blog-recognition software and has spidered out to pick up over 600,000 blogs. Not only has he been collecting the URLs and various global stats, he's been archiving the entries, too. You can download everything he has: the list of URLs, the language, blog tool and number of incoming links, the current HTML cache. If you want it all, you'll need to download around three gigs.

As he said in the talk, there's loads you could do with this data, "but I'm not imaginative enough to think of it". Nonetheless, he's already found out some fascinating stuff. If his language-guessing algorithms are right, over 1% of icelanders have a blog; Poland, Brazilians and Iranians love 'em, but most of South America and Spain are nowhere to be seen.

He's begging people to do something with this, and yet I haven't heard a single mention of it on the blogs I read. Hell, there's even an XML-RPC interface. How much more meme-worthy can you get?

I guess it's a crowded market, what with blogdex, technorati and organica. But this is an academic project, and it's open. Raw, crunchy downloadable data. At the very least, I bet the Internet Archive people might be interested in a constantly updated longitudinal archive of blog. Hell, maybe Google will slip them a few bucks for it.

2003-07-10»

More trailing-edge con watching. I'm spending more time observing the audience than the speakers, trying to overhear conversations that will give me a wider view of what's happening here. You can't keep track of everything, though. I'm most conscious of what I've missed. I've lost contact with the "whither open source in the enterprise" attendees - apart from Chris diBona, who is asking business model questions everywhere. The only bit of the corporate world I've heard about was various agog folk talking about a blistering Morgan Stanley talk on their internal Perl use. They described an amazing CPAN archive they maintain which has all the previous versions of the 280-so modules they use, all versioned so that their systems always use mods that are guaranteed to work with it. Plus all kinds of scary international desktop synchronisation over AFS. "They maintain a Perl system for 15000 programmers that we can't keep going for ten", said one guy.

I missed the Ruby folk too, much to my annoyance. They were a bit the underdogs at the conference, but everyone liked them. Most of the guru-level Perl coders admitted to messing around with Ruby for fun. You can see a few Ruby ideas percolating into Perl6.

One of the big themes for me was hearing the Perl guys wanting to help out everyone else, whether the other languages wanted them or not. That fits in with what's best described as the irrational exuberance of the Perlees. They run around like big slobbering St Bernards, knocking over the quietly studious Python guys and barging into the BOFs, barking and licking people whenever they found them. They really, really want everyone else to have a CPAN, for instance. That's one of the aims of the freepan project.

Freepan, along with FIT and YAML, is a Brian "ingy" Ingerson project. Looking back through the archives, he's always been fairly ubiquitous at OSCON, but he was very much in the epicenter of my OSCON this year. He's not necessarily the brains behind every good idea here, but he's usually a degree of separation away from it.

I think the next few weeks trackings will all come from here.

2003-07-09»

I keep being late for things. Here are my notes on the second half of Ward Cunningham and Brian Ingham's FIT talk, and the second half of Damian Conway's talk on Perl6 (aided by Larry Wall who heckled from the back using a scary "voice of god" microphone). I think I'm beginning to get Perl6, but it still does feel like I've stumbled into a shiny white room with bits of Ruby, Objective C and Python smeared up the walls and all of the Perl guys giggling in a corner.

FIT, though, looks very tasty, and social software if ever I saw it. It's unit testing meets wikis, which means that using it is like renting room in Ward Cunningham's head. I wonder a little if it doesn't require a bit too much futzing around in HTML textareas, but that's easily fixable with a front end.

I did meet up with esr. We didn't kill each other. It was touch and go for a bit though. Thankfully, it turns out you can block a lot of killer martial arts moves by holding a little baby in front of your face.

2003-07-08»

I'd tell you about the monster State of The Unions speeches, but I spent most of them either ferrying a cranky child around, or background-coding a blosxom moblogger for quinn and I to picture-o-gram the rest of OSCON. It's in no fit state to show anyone, but if you're interested, the program is really just a punched-up version of stripmime, some old code that ripped attachments out of mail so you could IMAP or POP them faster. I don't multi-task welll, so I'm going to have fun refactoring it later to remove all the lines that say $copyfiletolarrywallsbeingquitefunny = $doesadaneedachange++ and so forth. (Yoz has been doing the same sort of thing with Moveable Type if you want proper code from a proper programmer).

Larry Wall was funny (and Ada, generally speaking, did need a change). I'm always unsure what new technical details one can pick out of these talks when so much is available online. I listen mainly for the personal stories, the emotional inflections you would otherwise miss. Wall, in his gentle way, touched on how he'd sacrificed a fair bit of his career and his mortgage to work on Perl 6. He's just come out of hospital for ulcer treatment and it doesn't sound unrelated. He wasn't asking for pity, but he seemed relieved to announce that Damian and he had largely finished the Perl6 core language design.

I'm hoping to find my aha! moment with Perl 6 here. I think a lot of the Perl mongers are too. I heard a bunch of British Perlies cheering the new name for the Perl5-on-Parrot compiler, "Ponie". As in the cockney rhyming slang, "pony and trap". "There are many reasons for calling it Ponie", said Larry, "none of them good."

Guido mostly replayed his EuroPython keynote, which apart from the junking of the mooted Python equivalent to C's (a?b:c), was straightforward, uncontroversial and reliably, Pythonically, dull. Guido said he'd strip down the language even more, if he knew how. Lots of pictures of his new son, which seemed to me to be perfectly in order.

We ran away from the talk with a klaxxonning Ada after Guido, and only skipped back in to hear the final talk on the State of Linux from Ted T'so. A fair bit of discussion about how 2.5 is more Java-friendly, although the Linux guys still insist it's all Sun's fault for their approach to threads (and T'so did manage to slip in a comment about "Write once, run screaming"). A bit of snippiness, too, aimed at Eric Raymond's CML2 venture.

I'm actually meeting up with esr tomorrow. If he doesn't shoot me dead in the first five minutes, I'll hazard asking him about that.

2003-07-07»

We didn't make it to Damian Conway's talk. Ada was fine, but had sufficiently tortured her dear mother by rising early that by 4AM it was all we could do to slide into a motel just halfway up the six hundred miles. We ended up pulling into Oregon around seven this evening.

At least being out of of the Net gave me a chance to catch up on some writing. It's also a good way of keeping me from my usual procrastination-through-overresearching gig. For instance, I have recently been ploughing through a huge pile of academic economics research on ebay which I'm belatedly realising won't quite fit into a 1500 word how-to-sell-online-for-novices piece. Worth it if you're looking for a few white papers for your pez dispenser futures consultancy business, though.

As is traditional with long road trips, we listened to too many show-tunes.

2003-07-06»

I'm off to Portland, Oregon in a few hours. We'll be driving through the night, because Ada (pictured below eating my tie) isn't a great traveller, but happens to be a member of the US Olympic Sleeping-At-Night Team. We'll see.

It's for the O'Reilly Open Source Conference, which I'll try and document as much as I can.

I think this may be my last O'Reilly conf for a while, because I'm beginning to feel like a Deadhead, following them around like this.

Also, it's really about time I settled down and did something, instead of zooming around talking to other people about the somethings that they're doing. Or at least I should write much longer things about other people doing something. Stuff with indexes. Because if you don't write things that have indexes, you're nowhere.

Part of that search for a something is leading me to arrive, unshaven and sleep-deprived, having spent ten hours in the echo-chamber of my daughter's screams, to head straight for Damian Conway's 8.45am Monday talk, "Inside ~damian/bin".

I have this crazy idea for a book, and Damian's bin directory is just the thing to convince me it's a stupid undertaking. Or get me far too excited about it, one of the two.

After that, I'm just praying I don't collapse and sleep through the rest of the week, and miss Ward Cunningham's talk on the unit-testing for users, fit.

The rest of the time, I think I'll be moving in an unstructured haze between talks, ADHDing my way through the conference as usual. At Emerging Tech, I jumped around like this, but had this terribly embarassing experience of leaving every talk just as Tim O'Reilly was dropping in, leading him to deduce I'm sure that I was spending my time melodramatically storming out of every event. It didn't help, I don't think, that on the one occasion I was stable and sitting down when he entered, I'd just stolen Jeff Bezos' chair. Like Jeff Bezos needs a chair.

Anyway, anyway, come up and say hello if you're there. I'm the taller, dorky-looking gentleman above.