Tom Morris

Via PZ, Karen Heller at the Philadelphia Inquirer has an excellent article up about The Franklin, Philadelphia's science museum, pointing out the high costs of entry (admission is $23.25 per person for the over elevens) and exhibits filled with vacuous exhibitions that aren't high in scientific quality.

I have to say, one of the things I absolutely adore is the fact that London museums and galleries in the last few years have been encouraged by the government to reduce their entrance cost to that most wonderful price: free. The British Museum, the Natural History Museum, the Science Museum, and many more are now free, as are plenty of art galleries. You can now visit London, see an Archaeopteryx, ten Pablo Picasso works, an eighteenth century orrery, original Magna Cartas and Gutenberg Bibles without paying anything beyond the cost of a few Underground tickets. This is undoubtedly a good thing.

That said, if libraries are any measure, we need to be on guard against the kids-need-dumb-shit brigade. That may be so, but they don't need it in the museums. All those clever interactive doohickeys have a place: it's called the Web. I don't see the point of travelling some distance to go to a museum to find oneself interacting with a screen when you can do that at home. Me? I blame all these targets. The measure of a museum is simple: does it house objects of intellectual, historical, scientific, artistic and pedagogical importance and interest? That is what makes a good museum. Not how many people go through the door. Not how many press launches and junkets they have. Not media coverage. We should have art galleries because art itself is important, not because - as one government minister stated - consumption of art by ill people can often make them better. We should have museums about science and technology because those subjects are important, not so we can sell crap in the museum shop or increase the Sats results.

Measure museums as museums, not as subdivisions of hospitals, social services departments and job centres. There's no point getting a thousand more people through the door if there's nothing of substance to see.

Glad to see that Microsoft is embracing Ruby, and the open source community attached to Ruby. Yesterday, I downloaded Mono and managed to get IronRuby working on my Mac.

I got a bug report for Rena in the other day (ticket #2 in the bug tracker). It was a really simple bug that basically said XML was being serialized differently in JRuby than in the standard C Ruby implementation (MRI). I don't know why. It could be my code is being interpreted differently by JRuby or it could be that JRuby takes my calls to REXML and handles them differently - using native Java XML libraries, maybe. I didn't investigate. I'm not bothered what the difference is, so long as it works. I changed the test code so that the test passes and marked the bug as resolved. The next thing I did was added a new Rake task called 'jspec', which is the same as the 'spec' task, except it runs the tests through JRuby.

The tests take a lot longer in JRuby. On the standard interpreter (1.8.6 on Mac OS X, universal-darwin9.0), Rake says the test suite takes 0.516497 seconds, while on JRuby Rake says the tests take around 3.711 seconds to run. Time gives a different result to - on the standard interpreter, the tests take 4.968s (real), 2.365s (user) and 0.422 (sys) while the the JRuby tests take 13.718s (real), 8.789s (user) and 0.724s (sys). This seems to parralel what Wikipedia says about JRuby performance, namely that the interpreter is slow but it's much better on the server because of JIT compilation. If you are interested in performance, do read The Great Ruby Shootout. It goes into more detail than one could ever want about interpreter performance.

As for IronRuby? I managed to get it to work, and I could pull up an IRB shell without too much fuss, and I can run simple scripts, but as for being able to load gems and run tests (etc.), it isn't ready for that as far as I can tell. Still, it's looking pretty cool. I had to do quite a lot of dark magic to get it to work (grab old versions from the repository, apply weird patches and so on), and am somewhat astounded that it works. I'm tempted to not bother trying to run IronRuby on my Mac and just give in and do testing on my Windows machine which has actually got an official .NET implementation. Certainly, when it's possible, I'll be adding an "ironspec" task to the Rakefile so we can make sure those on the Microsoft side of the aisle can run the library.

Then, of course, we've got Rubinius, which is the next implementation I'm planning to test. They also seem to be operating in a nice, open way. I'm hoping that if library authors make a point of running their tests on these different interpreters, we can both make our code more portable, and help those developing the implementations.

I'm really excited that I can write a piece of software in a really high-level 'scripting' language that can now run in an ever-growing number of different implementations. (I reject the whole programming/scripting language divide as irrational - it's now just an arbitrary line with little use - especially now we have more and more abstractions, virtual machines and interpreters that run on virtual machines).

From Norm Geras, I see that a man called Jonathan Hoffman has published a report entitled Antisemitism on Guardian 'Comment is Free'. It's basically a report containing a bunch of posts containing a variety of idiotic and anti-semitic statements, and some poor moderation decisions that accompany them. And before we continue, there are plenty of very idiotic statements in the comments, the sort that you can imagine finding on many political blogs.

But some of the presuppositions that are stated in the opening pages of the report I strongly disagree with. It is this from the third page: Note that the advocacy of 'one [secular] State' is antisemitic. Says who? I think that secularism is a universal principle and should be applied in every country. According to Hoffman's standards, that makes me an anti-Semite. I think Israel should be a secular country - as I think the United States, the United Kingdom, India, China, Saudi Arabia, Iraq, Germany and every other country you name. I wish that Britain would finally hurry up and disestablish the Church of England, and I wish Iran would become a secular, liberal democracy. Does that make me anti-Anglican and anti-Islamic to an offensive extent? Not at all. I'm opposed to all forms of non-secular government, because I think that it necessarily involves the coercion of those people who do not subscribe to the established religion - whether that means forcing people to fund schools or churches which teach religious dogma, or forcing them not to practice their religion or to not criticise religious beliefs which they disagree with.

I disagree with the conclusions of the report too. I do not think that newspapers need to have comments - it makes little difference whether comments are pre-moderated or post-moderated (ignoring the impracticality of pre-moderating comments). I think that Comment is Free is mostly a steaming pile of shite and should probably be turned off.

I'm not a big fan of comments on blogs, and think that the Comment is Free comments are really the bottom of the barrel. If the Guardian wants to participate in the blogosphere, there's an easy way: tell the journalists to stop reading the comments and start using Technorati and Google Blog Search, as well as referrer logs, to hunt for reactions on blogs. Then if you get particularly interesting threads appearing in the blogosphere, link to them.

I really dislike the Dewey Decimal System. I used to think it was a brilliant system, but I've spent too many years in well-organised scholarly libraries (my college library uses Library of Congress classmarks and a home-brewed classification system) and gotten attracted to non-Dewey systems. My local public libraries use Dewey, and it is constantly irritating to me.

Currently, the 1xx region is for "Philosophy and psychology", but within that it's a mess. Psychology (15x) and "Paranormal phenomena" (13x) stick out like sore thumbs. There's an easy way to fix this. Move them to the end of the 1xx range - have Psychology as 18x and Paranormal phenomena as 19x, and shuffle all the philosophical topics down into the first seven ranges. If you were up for more radical change, then perhaps take Psychology out of the 1xx range and put it in the 3xx range - the social sciences. You could do that by compressing 35x and 36x.

In my local public library, there are a bunch of books on philosophical topics, but they are punctured with large amounts of cod psychology (of the Men are from Mars, Women are from Venus type) and paranormal claptrap. The Dewey system's 1xx range was probably fine for an age of Spiritualism and Phrenology (the latter getting it's own third-level category: 139) but in times where the cultish nonsenses are far more political and explicitly religious, the placement of 13x really does stand out. Similarly, now that psychology flourishes itself as a science, rather than as something of a latch-on to philosophy, it seems very strange to have the two mixed together.

This applies just as much to Library of Congress though - BF (Psychology) sticks out like a sore thumb in the B category. For religious topics, I'm a big fan of how LC do it though, and I could probably tell you down to a pretty good guess where you'd find stuff in the latter part of the B range.

In the end, it kind of makes you want to do it how some German libraries do it - have no categorisation scheme, and just put the books on the shelf based on publication date. Bring on the Semantic Web!

It's been most interesting watching the rather strange larval development of the various aspects of the Mozilla suite (formerly Netscape Communicator) into the separate pieces. I finally tried out Sunbird today.

Most of it isn't particularly interesting: it's an iCal client which can use the CalDAV protocol to subscribe to remote calendars. But there is one thing which makes it amazing: the Provider for Google Calendar plugin. How does this sound: a free, open source, cross-platform calendaring system that syncs with Google Calendar, and store those events for when you aren't connected to the cloud? Cool!

If I were in charge of the Google Calendar team over in sunny Mountain View, I'd get some of the engineers to start working on Sunbird - getting deep integration, cleaning the UI up a bit and so on. And Apple should really add a plugin architecture to iCal. Hopefully Sunbird will be a catalyst for nice lightweight, cloud-aware calendaring apps, nailing another few nails into the creaking coffin of Microsoft Outlook.

Last night, before going out for post-conference merriness after OpenTech 2008 (a really cool event, btw), I took part in a bit of a key-signing party. Key-signing has been made fun of on XKCD (leading to reader confusion). While I was there, David McBride showed me a tool on Debian Linux called "Caff" - it's up on the web here, described as a CA (as in Certificate Authority) "fire and forget" tool.

I looked at it, but it was a jumble of highly mind-bending Perl code, so I decided to rewrite it. In Ruby. For the Mac. After a lot of frustration, it's done. And I present KeyRub. KeyRub takes a space-separated list of GPG short-hexes (mine is "A6A4F54E"), goes online, grabs the key, shows you the fingerprint of the key, then presents you with the prompt to sign the key. If you choose to sign the key, it then makes an encrypted version of the key with your signature available for each of the listed e-mail addresses and mails them out. This is a much quicker process than sending random data out and then waiting for a response, then signing, then publishing. It also means that the signee can choose to publish your signature only if they want to.

In addition, because storing e-mail server passwords in plaintext just sucks, I have used the "security" command line utility (which is a command-line interface to the OS X Keychain - and boy is it easier to use it this way than faffing around with Python or Ruby Objective-C bindings) to get my Gmail password out. Obviously, this won't work on non-OS X systems. And if this script wasn't built only for my own usage, I'd make it so that it would extract my Gmail username from the Keychain also. But that isn't likely to change too often. Hint: if you are building command-line-based scripts that interact with servers or APIs, use the Keychain on OS X. It rocks.

What is all this key-signing stuff about? Well, you should start using GPG. Even our supposedly liberal, freedom-respecting, Western governments are screwing around with our privacy, and GPG allows you to fight back. Use it to encrypt your e-mails and files, as well as do things like sign software (this is tremendously useful - if you make a piece of software, signing it means that others can redistribute it, safe in the knowledge that others can check its veracity). If you are in London, I am often at geek events, and am always willing to sign keys - my policy is I accept only government-issued IDs (passport or driving licence) with discretionary cultural name variability (ie. if your passport says "Thomas" but your key says "Tom", that's fine - if your passport says "Thomas" and your key says "31337h@x0rZftw!1!", think again).

Recently, there has been a fair bit of hype around a new Firefox extension called the Universal Edit Button. I strongly advise against using it. It misuses the semantics of the link element in HTML. And not just a little bit, like really the sort of bad usage that makes one's eyeballs spin like a slot machine type bad.

Get this. It uses rel="alternate", and type="application/wiki" - oh, but because that's not a proper MIME type, they ask people to use type="application/x-wiki". Because, you know, when you edit a wiki page, instead of getting an HTML document back containing a form, you get back an "application/x-wiki" document. I am not kidding about this.

There's a really simple solution to this - just make it so that it is rel="editable" (or a similar rel value). And why even bother having it in the head? Have it on the actual link to the editable version.

I strongly advise owners of wiki systems and other collaborative systems to not implement this until it stops sucking. You may want to go and suggest this on their wiki.

Just catching up with my newsfeeds, and I see this story from the W3C. It's an announcement of the "Common Web Language (CWL) Evaluation and Installation Incubator Group". Only problem is that it seems to be almost completely opaque to anybody outside of it. The XG charter describes CWL as a graphic language of semantic network with hyper node and is used to describe contents and meta-data of web pages in three different type of form such as UNL, CDL and RDF. Anyone understand that? I don't. I get why one might want to describe content and meta-data of a web page in RDF - but why do I need to use UNL or CDL? That was my initial reaction.

But then I read this XG Report about the Common Web Language. And, my, that is really quite bonkers. It's Esperanto-over-HTTP! Finally, we have people who are barmier than us!

From what I can gather, the UNL/CDL effort is trying to map a fairly large set of human language into a limited vocabulary which is easier for machines to translate, and then using RDF to attach metadata. Instead of sitting down and writing in English or Japanese or German, one sits down and writes in UNL. Then, the browser has a UNL reader built in which automatically translates the UNL into the relevant local language. In short: machine readable Esperanto with RDF and OWL annotations. Right, own up: who at the W3C has been smoking crack? At least it makes us "HTML, microformats, RDF, SPARQL and a bit of OWL" pragmatic pick-and-mix types look completely sane in comparison. I don't want to go all negative-nancy on anyone, but anyone want to bet that this will actually take off? Have I got it all wrong - and we'll all be writing CWL instead of HTML and XML in five years time?

Oh my, here is their example of the beauty of the UNL syntax. Here's the English: Long ago, in the city of Babylon, people begun to build a huge tower, which seemed about to reach the heavens. And here it is rewritten in UNL: begin.entry.past-- tim->long ago --plc->city.def --agt->people.def --obj->build.past--obj-> tower<- aoj--huge<- seem.past --obj->reach.bigin.soon--obj->tower --gol->heaven.def.pl

I propose a new rule of thumb, which I shall egotistically call Morris’ Law of Standards: however fucked up and crazy something is, someone, somewhere in a standards body is writing a parser, schema or proposal for it. CWL proves that. It really is machine readable Esperanto over HTTP. At least they (claim) to use N-Triples in their paper (it’s not, but don’t worry about it) - Esperanto re-expressed in RDF/XML would quite probably make my brain asplode.