[
Lists Home |
Date Index |
Thread Index
]
David Lyon wrote:
> Liam,
>
> I was reading this article...
>
> http://www.advogato.org/article/820.html
>
> so everything is looking good for an xml 2.0... with datatyping
> and other good stuff...
"So I'm with Jim Gettys here, let's drain the swamp. XML might not be
perfect for this, but it's better than today's nonsensical mess."
With regard to Liam's observation on configuration above. The Java world
has made heavy use of XML configuration and represents to my knowledge
the best available data point for any effort in that direction.
Although using XML is a handy way to get configuration /into/ Java
systems (once you bind to a tool like Digester or XStream), it's not
clear how beneficial that has been in draining swamps with regard to the
formats themselves. There are a *lot* of XML config formats out there.
In a Java project, dealing with a dozen different formats is typical.
Maintaining systems using scads of XML deployment descriptors and config
files is a pita, especially in a cluster. The current situation is known
to be a mess - it's not at all clear that XML is helping.
XML alone arguably offers too much flexibility for a configuration
language; what's needed are further constraints in the way Ant
constrains task coordination or RDF constrains relationships. Do check
out smartfrog [1] for an alternative approach to this problem (btw, one
of the people involved in smartfrog, Steve Loughran, is also a committer
to Ant).
Utter Speculation: I think one of the issues people have reading markup
for configuration is that XML structure drowns out the information,
unless you go for heavily attribute driven markup, which XML people tend
to frown on. XML seems to be clearer when things are the size of
paragraphs rather than sentences. Given XML's roots maybe that's not
altogether surprising.
cheers
Bill
[1] http://www.smartfrog.org/
|
