[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

RE: Another "Against the Grain" column on XML

• From: Frank Richards <frichards@s...>
• To: xml-dev <xml-dev@l...>
• Date: Fri, 10 Aug 2001 14:55:04 -0400

<< And frankly, doing so using an RDBMS has frequently felt like programming
a Turing machine. Sure you can do anything that way, and you know you can.
But it sure is the hard way. This is complex multidemsional __information__
and splitting it into 2D tables is work, and then getting it back in a
useful way is a major task for the server.

Perhaps I'm misreading your posting. I'm addressing my comments to points I
believe you are trying to make:

1. You implied XML is complex information for which two-dimensional
operations are inadequate.

Ken,

I've worked on three major text/bibliographic apps.

1. A science fiction bibliographic database, that not only catalogued the
collection,
but also indexed the works as historical chronicles on the universes in
which they were
set, and also tracked "litcrit" info about versions and rewrites.

Obviously this was a hobby effort but the hobbyists were librarians and
programmers.
It was many years ago and we chose a network database deliberately: even in
that long ago
day the RDBMS were much more polished and invested-in products, but because
we were
modeling a complex multidimensional network, it saved so much in
normalizing/denormalizing.

2. I published a legal full text database for a few years -- Absolutely
classic full text app.
Tag-aware full text indexing was the only way to travel. The full text tools
could handle titles and dates,
an RDBS just wasn't there on the full text.

3. Another bibliographic app. -- Indexing and abstracting. We used an RDBMS.
The project failed.





>>>
2. You implied the relational model means having to map data into a 2D model
when doing queries.

Some SQL implementations support multi-dimensional expressions. You can, for
example, run queries that operate on a data cube. You can also write star
schemas for multi-dimensional data sets. That's a common technique for data
warehouse design.
<<<

Wanna bet that it doesn't cause Fabian Pascal to foam at the mouth, and Codd
to claim
that it's not really relational?

>>>
What query tool do you use that supports multi-dimensional expressions when
querying or navigating through XML documents? Do you have, for example,
software that let's you write XPath with dimensions (e.g., reference the
dimension of a node-set or an axis)?
<<<

Two that have done good things for me in the past are Dynatext, and Poet's
OODBMS.
No neither does everything. In particular, neither scales the way any of the
big RDBMS do.
Both do allow minor changes to the data model without causing heart attacks.
But that I claim is largely a matter of time and money -- give me half the
years and a tenth the
R&D money that has gone into Oracle (since day one) and a lot of product
maturity issues will go away.

Also, I see two different applications here -- finding information in a text
corpus, and
keeping track of information about that text corpus. Neither seems to me to
fit well with
an RDBMS which is all about how many copies do you have of ISBN 0-123-....


Frank

• References:
  • Re: Another "Against the Grain" column on XML
    • From: Ken North <ken_north@c...>

• Prev by Date: Re: Another "Against the Grain" column on XML
• Next by Date: SAX-core proposal: class loading and META-INF/services
• Previous by thread: Re: Another "Against the Grain" column on XML
• Next by thread: RE: Another "Against the Grain" column on XML
• Index(es):
  • Date
  • Thread