Tim Bray on version 2 of the RDDL spec. Its fascinating to watch how incredibly important human language is in the XML world. Take something like the word "purpose". If you are marking up the purpose of something, what better word to use in an attribute or element tpe name than the word "purpose"? We like to establish one-to-one mappings between concepts and words. Unfortunately, human language does not facilitate this. Words that mean many different things depending on context are par for the course. If we used semantic-free tag names and "data dictionaries" the problem would not be so immediate but we would have re-invented EDI which would be deeply bad.
The architectural form concept from HyTime (and the SGML days in general). Addressed this problem of mapping names to semantics to an extent but unfortunately, inevitably, it did so at the expense of another layer of abstraction.
There is a universal law in there somewhere...

posted by Sean 3:01 AM

[Link] . . .

An ITWorld article on the subject of numbers and units of measurement with a detour down the strangely signposted roads on Ireland.

posted by Sean 2:24 AM

[Link] . . .

Web Developer is really nice. I've now got a toolbar with all the geek stuff that it either not possible or buried 15 levels deep in some menu structure. Have a look at the feature list.

posted by Sean 2:12 AM

[Link] . . .

I have updated and added stuff to my bookshelf page and uploaded the XML file used to generate it.
As I understand it, the ASIN number for a book on Amazon is its ISBN. Unfortunately, not all ASIN's that I dig out of amazon.co.uk will work when I construct the necessary URI to access amazon.com. Case in point On Nature and Language works but The man who loved only numbers doesn't but it is in amazon.com's US catalog here. Go figure.

posted by Sean 2:59 AM

[Link] . . .

I will be giving the keynote talk at this event in Cambridge at the end of the month. The talk is entitled "Modelling with XML - puzzles or problems?".

posted by Sean 6:03 AM

[Link] . . .

James Britt on Duck Typing and the origins of the phrase in the Ruby language. I think I first heard the phrase from Frederik Lundh. I latched onto it straight away because it brilliantly summed up Python's runtime typing. I first realised its power when working with Web Resources. When you call urlopen you get back a file-like object. The "file-like" bit intrigued me. What is a file? It is an object that has certain methods and attributes. The combination of methods+attributes is what is important, not the file-ness of its base type. In more philosophical terms (Kant I think), does the "file" abstraction have any intrisic properties or are all properties that make up file-ness extrinsic? If there are no instrinsic properties, can there be a sound ontological basis for static typing?

posted by Sean 2:28 AM

[Link] . . .