[
Lists Home |
Date Index |
Thread Index
]
Yes, I understand the complexity of using the content model
for that, and in X3D, we have some very complicated content
models as a result of that. Seems to me the Schema approach
is simpler to write. On the other hand, a heavily parameterized
DTD is also a really ugly thing (the X3D DTD approaches
unreadability). Should we have gotten rid of parameter entities
in XML? The DTD design gurus fought for them like tigers.
That side trip aside, again:
1. Not all Schemas are designed for generality.
2. There are use cases for mix/max and some of those
can as easily be satisfied by the content model. There
are also cases where min/max is easier.
3. Relational model examples don't necessarily apply to
document examples and XML is for the most part, a document model.
None of those are showstoppers or "proves the rule" arguments.
I don't think we'll discover one of those in this topic. We'll
get use cases and those will be particular, and we'll get generalizations
and those won't always apply to particular cases. It seems to
me the min/max feature is yetAnotherToBeUsedSmartlyButAtOnesPeril
features.
I am unsure if some are arguing here against min/max to support
RELAX NG or to improve XML Schema. That may not matter.
len
From: John Cowan [mailto:cowan@mercury.ccil.org]
Bullard, Claude L (Len) scripsit:
> I grant that in more cases than not, leaving the max
> occurrences infinite works, but wasn't the lack of
> being able to model a finite set one of the precise
> reasons for critiqueing DTD modeling?
It might have been an excuse, but not a reason: m,n-bounded
repetition can definitely be done in DTDs, though it is
even more annoying than in RNG, where f,f,f,f?,f?,f?,f? is
legal for f{3,7} as it is not in DTDs, where you must
say f,f,f,(f,(f,(f,f?)?)?)?.
|
