Internet Alchemy

Happy New Year to everyone who knows me or reads this weblog.

Here’s hoping that 2003 is going to be fantastic and that I can forget that 2002 ever happened (except 6th August!).

I have finally written up a list of 11 issues I have with the current RSS 1.0 specification.

Continue Reading »

I’m experimenting with adding geographic metadata to this site. The home page now has additional meta tags to enable geographic searching at GeoTags and GeoURL:

<meta name="ICBM" content="52.422400,-0.803000" />
<meta name="DC.TITLE" content="Internet Alchemy Weblog" />
<meta name="geo.position" content="52.422400;-0.803000" />
<meta name="geo.region" content="GB-NTH" />
<meta name="geo.placename" content="Rothwell,  Northamptonshire, United Kingdom" />

These tags can, in theory, be used to pinpoint the geographical focus of this weblog, whatever that means. It might make more sense to add it to my FOAF.

I’ve added an experimental search box to the sidebar to enable searching of postings to this weblog. It currently only searches the main content of each posting, not any comments that people have left.

A new phrase bumping around the Wiki community is WysiWiki which denotes a wiki that has a live, editable interface rather than having to press a button and type strange forms of markup into a text box. Pepys was my contribution to this space, and now I have another: WysiWiki.com is a website for discussion and debate about WysiWikis. It’s organised as a Wiki naturally and it’s not intended to be tied only to discussion of Pepys but all kinds of more user-friendly wikis.

This is something I knocked up a couple of nights ago to learn more
about N3.

It’s a Wiki that uses N3 as its underlying format. It’s written in
Perl, runs as a CGI and uses RDF::Notation3 to parse the N3. [source code]

Each wiki is bound to a particular namespace and when the N3 is
rendered as HTML for browsing, any resources in that namespace are
linked to their wiki page automatically. If the resource has no wiki
page defined then a hyperlinked question mark appears after the name
indicating that you can create the N3 definition by following the
link.

Each page can be viewed as HTML, RDF or the raw N3.

Here’s a demo of it working.

Create new resources by browsing the the appropriate page, e.g.:

http://internetalchemy.org/animals/Frog

http://internetalchemy.org/animals/Toad

http://internetalchemy.org/animals/Sandpiper

I chose RDF/N3 over RDF/XML because cvs integration will work better
when I get around to adding page histories.

Might find some use in building vocabularies collaboratively.

It’s only hacked together really, so expect bugs, gotchas and
unexplainable behaviour.

Another potential element for identifying businesses came to me last night: tax reference, the company’s reference number for sales tax / VAT. May or may not be a core informational element. I don’t know yet.

James has suggested that a good place to start with defining a business description format is a Dublin Core-like set of informational elements that can characterise a business.

That set me thinking about what characteristics are exhibited by Dublin Core that sets it apart. My view is that the Dublin Core
element set consists of 15 deliberately vague elements. For example,
date represents a date for a resource. What kind of date? What format?
Whose calendar? Dublin core doesn’t specify any of those, and it
doesn’t really care – it’s up to the author to interpret the meaning
in the context of the resource they are describing. The DC qualifiers
can add more precision (created, modified etc). This is, I believe,
the secret of Dublin Core’s success. Because they are not
proscriptive the elements are applicable across a wide spectrum of
scenarios: from xml schemas to rss to gutenberg etc. Also, by keeping
them simple but extensible the first set of elements could be adopted
quickly by authors.

Additionally, the Dublin Core elements generally describe invariants about a
document: creator, publisher, identifier, language etc. It’s another
factor in Dublin Core’s success – maintenance of the elements by an
author is rarely required.

So, we have to ask ourselves, what are the CORE elements of
description about businesses and organisations and can we abstract
them to eliminate cultural differences?

Currently I don’t think address information is core to identifying a
business or organisation. It varies more rapidly than the company name.

Anyway these are the concepts I came up with (some are dubious of
course):

• registration – the registration number of the company or organisation
• name – the name of the organisation
• type – the type of organisation (e.g. Ltd., Plc, Corp., Gmbh., charity)
• trading name – an alternative name the organisation is known as
• classification – the business classification (NAICS, SICS etc)
• stock symbol – the stock symbol the company trades under (MSFT etc)
• country – the country the organisation is registered in

Alex Chaffee: I have a feeling that Erich Gamma
writes entire applications solely via right-clicking. No
typing, not even any left-clicking. Pretty much any
interface *other* than right-clicking seems like an
afterthought in Eclipse

Something that struck me when browsing WordNet for types of
premises was that shop types (such as toy shop or computer store) are
best characterised by the types of goods they sell. Pretty obvious stuff, but
it led me to think about how businesses are classified. Common codes such as NAICS
or SICS group the types of business in a loose hierarchy e.g.:

AGRICULTURE, FORESTRY AND FISHING
|-.	AGRICULTURE AND HORTICULTURE
  |-.	AGRICULTURE
    |- CATTLE FARMING

One of the key topics of research I delved into at Calaba was in the realms
of what we called multidimensional classification (cropping up on the web these
days as faceted classification). The key to this kind of classification is
decomposition of monolithic classification hierarchies into multiple orthogonal
hierarchies. The key example we always used to illustrate this process was a
hypothertical Yahoo category hierarchy that ran Sport / Soccer / Italy / Teams /
AC Milan – a category for websites about the AC Milan football club. The Teams
category is repeated under each country name in the directory. If one wanted to
include an English website about the football club, one would have to introduce
an English category at some arbitrary point in the hierarchy requiring even more
duplication. What we did at Calaba was to split these hierarchies out into their
orthogonal components: subject, location and language. The user
could browse each axis in combination or independently.

Anyway, the same concept can be applied to business classifications. Separate
out the function of the business and the subject of that function. Typical
functions might be Retail, Distribution, Manufacturing or Publishing. These are
then qualified by the types of products or services that are retailed
distributed, manufactured or published. e.g.

<businessType>
  <Manufacturer>
    <product>Cheese</product>
    <product>Butter</product>
    <product>Yoghurt</product>
  </Manufacturer>
</businessType>

The larger the business, the more types of business it performs and the wider
the range of products and services it deals with. This scheme is extensible
in a distributed manner. When someone invents a left-handed hammer then the
various businesses in the tool supply chain can just add it to their descriptions instead
of waiting for a government department to add it to their industry classification scheme.