On Wed, 16 Sep 2020 03:25:54 +1000
John Gardner <gardnerjo...@gmail.com> wrote:
> That's what I mean: there isn't always an obvious distinction between
> data and metadata. For structured HTML-like documents, it's common to
> see attributes used for anything not seen by readers (i.g.,
> "metadata"). For general purpose data serialisation, things aren't as
> clear-cut. In both cases, the decision to use attributes is still an
> arbitrary one.
That phenomenon has a name: accidental complexity. As opposed to
inherent complexity, where the complexity is inherent in the problem,
accidental complexity makes things more complex without adding any
power to solve the problem.
For years I looked for a general tool to import XML data in to SQL
databases. When I finally tried to write one, I realized there was no
automatic way possible. The data I was interested in was tabular, but
XML has two equivalent attribute-nesting syntaxes (attributes and
subordinate tags). To import anything, it's first necessary to map the
XML structure onto SQL tables. That pretty much eliminates anything
"automatic".
> > this *isn't* the same to me: attributes are for metadata and tag
> > contents are for data
That's right. If XML attributes were always used to describe the
*tags*, and not the data, there would be (almost) no problem. At least
we'd know where the data are. But it's not uncommon to see XML like
<record col1:"a", col2:"b">
<col3>c</col> </record>
Like so many might-have-beens, the XML community could have adopted a
standard for relational data, ideally one that didn't swamp the data
with markup. Alas, history is written by the victors, not the best
informed.
--jkl
