Re: WWW/Internet 2009: 2nd CFP until 21 September

Walter Mitty wrote:

>>> Anyway, I'm interested in whether XML falls under the topic of machine
>>> representation of the data and is therefore neither compatible nor
>>> incompatible with a relational view of data.

Is XML a 'machine representation of data'?

XML is a mapping from labeled trees with Unicode-valued nodes to Unicode strings. It maps trees to strings. Strings are closer to machine representations than trees, so yes, XML can be said to be a possible stage in machine representation of tree-structured data.

Furthermore, ordered labeled trees as expressed in XML are closer to machine representation than e.g. the relations of the relational model. So data described in terms of ordered labeled trees is indeed closer to the machine than describing them in a relational data model.

>>> Or whether XML is an alternative tro the relational view of data, and
>>> therefore one that should be compared with the relational view of data
>>> with regard to benefits and drawbacks.

Yes, it is an alternative: it was intended as a basis for text markup languages, which do not have very comfortable 1NF relational expressions, as far as I can see. It was not intended as a replacement for relational databases. Once it existed, it became popular for purely structural data descriptions, which compete with relational representations to some extent. Erik Naggum describes this well:

  http://groups.google.nl/group/comp.lang.lisp/msg/9a30c508201627ee?hl=nl

[...]

>It went something like this: In FORTRAN, data definitions were scattered
>all over the program, in FORMAT statements. In COBOL, the definitions
>were at least gathered at the front of the program, in the data division.
>Soon afterwards record definition libraries began to be accepted in the
>COBOL world. (BTW, I was never a COBOL guy). This enabled lots of
>programs to share record definitions. Finally, databases that contained
>their own schema, allowed data to be self describing.
>
>I'm wondering if the people who invented XML didn't know that this work had
>been done before, or if they regarded the work on databases as worthy of
>being ignored.

Well, this criticism can be applied to a lot of work on XML and RDF and to many of their applications. There is a large overlap in scope. But there is also a true difference. XML is for documents: structured data, stored and transferred piecewise. The relational model is for databases: large repositories that assumed to contain all information of interest. Of course you can regard a database as a big document, or a structured document as a said database, but calling them documents means we're doing different things with them than calling them databases. So the average use case is different; but languages such as XML Schema and XQuery attempt to bridge that difference, and I believe your remark applies to them to some extent.

>In any event, they seem to have reinvented the hierarchical model of data.
>This keeps happening. Next thing you know, we'll have somebody in this
>forum telling us that Nelson Pick got it everything right. We've already
>been down that road, but it can happen again.

Definitely. This newsgroup has been down that road, apparently, but but I haven't. Against the description at

  ftp://ftp.software.ibm.com/software/data/u2/pubs/whitepapers/nested_rdbms.pdf

most of the sneering and other types of fallout I tend to see here regarding non-1NF models completely misses the point.

-- 
Reinier