Lists Home |
Date Index |
Basically, I concur with Joseph. One can use attributes
for fields, certainly, and that will produce a compact
instance. It will also result in a much simpler schema
should one have to write one, but the restrictions on
what one can express in attributes viz a vis types and
structure also means the schema is not doing very much
useful work. It can also limit the use of the instance
particulary with regards to extensibility.
One can consider the instance also in terms of broader
contexts. What I find more and more is that if an XML
instance is dumped at all from a relational system, the
odds are good that it isn't being roundtripped back to
the same system; it is going elsewhere, perhaps to another
relational system, or even to the browser. This can mean
that one has to provide more than the original table did,
for example, alternative representations of say timestamps,
personnel numbers, and so on. In this case, one begins
to need the structural powers of elements. I seldom see
the relational data dumped precisely as it was used in
the relational table (and one has to account for nulls).
So best practice, meaning, this is not a one off,
tends to be as Joseph describes it. Elements are
From: Chiusano Joseph [mailto:email@example.com]
whether to use attribute or element based tagging here..
IMHO, this should not be an "or" type of choice - elements should be
used to represent data, attributes to qualify (modify) metadata. In
terms of english, one may think (in a very basic sense) of attributes as
adjectives, and elements as nouns.
Booz | Allen | Hamilton