[
Lists Home |
Date Index |
Thread Index
]
Tim Bray]
[
>I don't think any of them are cost-effective. If you're not 100% sure
what you're getting,
>use a real XML processor and the problems go away. -Tim
I won't go over the old ground again [1] but this is where the problems start
rather than go away for read/write XML applications.
If I use Perl or something, my app is simple, non-invasive to the
byte-stream, but possibly
wrong. If I use an infoset (e.g. I parse the source) my app is not simple,
invasive to
the byte stream but right.
Achieving non-invasivity + correctness only comes at the expense of simplicity.
For nuclear reactors and heart monitors, we need correctness. Sadly it
appears that
complexity is thus thrust upon us. All because, in my opinion, there is no way
to specify, up front and personal, that your XML data DOES NOT use CDATA
sections,
tags embedded in comments, internal doctype subsets, pi's in the epilog
etc. etc.
If I had such a thing in the XML rec, I could rest easy with my simple
Perl (actually Python :-) programs even in the face of possible upstream
changes.
Sean
[1] http://lists.xml.org/archives/xml-dev/200102/msg00584.html
|