Lists Home |
Date Index |
agree with Doug. Unless there is clear evidence that a feature is creating
a problem or
soon, don't close an extensibility feature. Wildcards, xsd:any, this
sort of construct,
are the real heart of the 'extensibility' of XML. Otherwise,
schemas are just a means to
validate. Even if there is overlap, it sometimes dicey to
turn a validator into a filter.
Sometimes one has to limit the depth of a recursion and when we had the
could be specified (support varied but the intent was clear).
Otherwise, I don't think
so far as to say 'never use it' because it, like mixed content, proves to be
property of the data. I don't like to constrain
organic properties to suit the tool.
bites. To put it another way, in a domain of attractors, you
can rely on
rules on the limit and fixed types, but the strange ones remain
light hand on the tiller.
Personal opinion: Recursive structures can be among the
most valuable structures available in XML, especially with regard to generic
<section> (and similar) elements. Putting a cap on maxOccurs and
adjusting as needed going forward can be a good idea.
Frankly, nothing in your arguments below justify the
statement "don't use recursive constructions." You do make a good argument for
restricting maxOccurs. An awful lot of the discussion is related to
specific "possible" conditions that can lead to problems. It is very
subjective. If recursion is not good for what you do, don't use it. But
telling others that they shouldn't when it is an important, proven part
of their processes just doesn't float. Quite a few major XML players are
proponents of recursive sections, based largely on experience. Of course,
quite a few are of the opposite opinion as well. :-)
I've written XML tips illustrating both potential
strengths and potential pitfalls of recursive sections, but would need time to
rewrite them here (since the online journal I wrote them for owns the
copyright to the exact phrasing of the tips). The point is, the value of
recursion depends largely on what you are trying to