[
Lists Home |
Date Index |
Thread Index
]
HTML parsers don't necessarily "fix" tag soup when they do their best to
render it. In the XML case, the well-formedness (w-f) constraint
requires that at some point errors are actually fixed. So the
application of "Postel's law" seems to be about WHERE this happens.
One of the problems with Tim Bray's (excellent) examples is that they're
examples of corruption. The well-formedness constraint is nowhere near
sufficient to catch all corruption; also, it's relatively easy to apply
heuristics which make w-f unnecessary (for that). For instance, if
there's at least one closing tag at the end
<trade>
<ticker>IBM</ticker>
<amount>100</amount>
then we don't have the main problem, even though w-f is still violated.
Adding the second missing tag still wouldn't guarantee whether or not
the figure used to be "£100" but the £-sign was thrown away by an
intermediate 7-bit transport.
It seems to me that the "woman in Baghdad" example is a key one. Do we
expect XML to be written (i) by humans and (ii) in situations where they
can't check w-f before shipping it out, or can, but their checker is
broken? If yes, it still doesn't mean that the check can't be done at
the ingress to the "XML-strict part of the system" somewhere else, but
it does mean that the original author may not be available to fix it up.
There may, however, or may not, be a non-realtime feedback loop that is
regarded as sufficient. Isn't this a business-workflow issue rather than
one of technical interoperability?
Notably, however, the only reference to "human" in the Goals section of
Rec-XML <http://www.w3.org/TR/REC-xml#sec-origin-goals> is that "XML
documents should be human-legible": not -writable.
Greg.
|
