Re: XML.org implementation (Packages)

["james@xmlTree.com" <james@xmltree.com>]

Hi

> We also looked again at classifications for registered items   --- this I
> will address again in a follow-on topic regarding packaging..etc..
> The general conclusion at the end of the conversation was to use the
> existing homegrown XML.org "application type" as another axis for
> classification with the knowledge that this will need to expanded or
> replaced over time.
>
> In reviewing the reggrep list --- I came across the link to xmltree.com
> again --- they use the Dewey system as one classification axis --- do we
> know did they license this or are they just using it?  I thought the problem
> with us using Dewey was in part because they didn't want people just copying
> it on the Web?

xmlTree has moved away from the Dewey system of classification for the following reasons:

1) It's difficult to arrange licensing terms.  I wrote to Joan Mitchell of OCLC in July
1999, and then again in August 1999, about how far one could use the Dewey codes without
license infringement, and each time she wrote back and said that it was being considered.
It seems that it's OK to use the top 3 levels, but this is woefully inadequate for
internet resources (IMHO).  For example, there is no subject category at these levels for
wireless usage, so where would WML 1.1 or i-Mode or HDML (assuming they were all XML) be
classified? I know that they are working to make the classification titles simpler, but
they have to bring the libraries along with them, so changes to the underlying structure
take longer.  Please confirm my understanding of the licensing rules with her, because I
am not entirely sure and don't want to misrepresent OCLC or Forest Press.

2) We are classifying a lot of WML content, and location classification is important.  One
short-coming we see in Dewey, DMOZ and Yahoo is that subject and geography (as well as
content type and language) classifications are shoe-horned into one structure.  We believe
that a submitter should be allowed to classify resources in multiple ways, along
orthogonal axes, and that dependencies should be removed from classification structures.
What this means is that there should be no geographic reference in a subject
classification - the submitter should choose a subject classification, and then a
geographical classification if relevant.  So we've built a structure that allows
submitters to do this.  And an additional benefit is that the classification hierarchies
become simpler.  The xmlTree classification has an XML interface (of course!).  The next
step is to add a hook into WordNet (an open semantic network), so that a thesaurus can be
used to improve browsing.  I am hoping to meet the WordNet people in a week's time.

kind regards,
James Carlyle
www.xmltree.com - directory of XML content