[
Lists Home |
Date Index |
Thread Index
]
At 02:27 PM 6/9/2003 +0200, bryan wrote:
>I was hoping someone could send me some pointers to interesting parts of
>the xml schema sucks perma-thread with special attention given to
>contested areas of the specification leading to vendor incompatible
>implementations.
I'm not prepared to send all the pointers, but I think these are the
biggest items:
1. Part 1, which governs complex types, is very difficult to read. It uses
a descriptive mechanism that lacks both the precision of formal notation
and the easy readability of well written prose. I have a hard time finding
things in the spec.
2. The XML representation is hard to read and write. The verbosity is also
a problem - it is hard, for instance, to show examples in W3C XML Schema
notation in the text of an article or chapter.
3. The type hierarchies are not simple - the distinct hierarchies for
complex and simple types and the distinction between elements and complex
types result in a more cluttered type system than that found in most OO or
relational systems. This is not the kind of type lattice that a good
datahead would normally design.
4. There is no normalized form that can be accessed as XML, and many ways
to express the same facts, so schemas are hard to query or manipulate with
tools.
5. For many document-oriented applications, especially those which do not
need types, W3C XML Schema has few advantages over DTDs.
6. For applications which need only the addition of simple types to
structures, W3C XML Schema is a big step in additional complexity, most of
which is not needed for these applications.
7. The complexity has important costs. Implementing W3C XML Schema is
beyond most small software development groups. The implementations of
larger vendors are not as compatible as users would like. There is
follow-on complexity for other specifications that must support W3C XML Schema.
Despite these flaws, it is the only schema language besides DTDs that has
been widely adopted in commercial tools. I think many people use tools to
generate their schemas, which makes it easier to author them and makes the
XML notation somewhat less of an issue. Tutorials are now making it easier
to learn for people who can't read the spec. I don't think it is an ideal
schema language, but it is widely supported, and it does handle most cases
well enough that it is unlikely that vendors will discard it.
Hope this summary is useful,
Jonathan
|