DTD or XML Schema. Which one is better?

Question

What are the pros / cons in DTD and XML Schemas (I'm not even sure what the official name of the latter is!)? Which is better? Why do we need two ways to do the same thing?

Edit: I found this in an article I was reading, which is what prompted me to ask the question:

Why W3C XML Schema Language?

The W3C XML Schema Language is not the only schema language. In fact, the XML specification describes document-type definitions (DTDs) as the way to express a schema. In addition, pre-release versions of the JAXB Reference Implementation worked only with DTDs -- that is, not with schemas written in the XML Schema Language. However, the XML Schema Language is much richer than DTDs. For example, schemas written in the XML Schema Language can describe structural relationships and data types that can't be expressed (or can't easily be expressed) in DTDs. There are tools available to convert DTDs to the W3C XML Schema Language, so if you have DTD-based schemas that you used with an earlier version of the JAXB Reference Implementation, you can use these tools to convert the schemas to XML Schema Language. http://java.sun.com/developer/technicalArticles/WebServices/jaxb/#binsch

I guess I would like examples that illustrate why XML-Schema is better (if it indeed is).

Solution 1

From http://weblogs.asp.net/rchartier/archive/2006/03/21/440782.aspx

• DTD's are not namespace aware.

• DTD's have #define, #include, and #ifdef -- or, less C-oriented, the ability to define shorthand abbreviations, external content, and some conditional parsing.

• A DTD describes the entire XML document (even if it leaves "holes"); a schema can define portions.

• XSD has a type system.

• XSD has a much richer language for describing what element or attribute content "looks like." This is related to the type system.

• You can put a DTD inline into an XML document, you cannot do this with XSD. This means DTD's are more secure (you only have to protect one bytestream -- the xml/dtd -- and not
  multiple).

• The official definition of "valid XML" requires a DTD. Since this may be impractical, if not impossible, you often have to settle for schema-valid, which is not quite the same.

For my part, it's pretty straightforward to write a validator for some XML if you have an XSD. I haven't seen this with a DTD, although I'm sure it exists.

Solution 2

A few years ago, there would be reasons to use DTD over XML Schema (it was more common or better supported by XML tools). Today, however, I see no reason to not use XML Schema instead of DTD : XML Schema is much more powerful.

However, XML Schema is far from being perfect (just try to read the spec or a book on XML Schema...) and many alternatives have been developed since then (Schematron, Examplotron, RelaxNG). These may have technical advantages over XML Schema, but XML Schema is so much more pervasive today that I see very few cases where an alternative would make sense.

Solution 3

XML Schema can perform more complex validations. For example if DTD can check if the datatype of an XML element is integer or string. Whereas XML schema can perform more complicated validations like if the xml element is a string starting with uppercase letter or a positve integer. Finally XML schema uses XML syntax and its a natural choice for development of web services.

Solution 4

There is also Relax NG — another powerful language for validating XML documents, along with Schematron and other technologies from DSDL. Relax NG is very simple and have human readable form — Relax NG Compact that allows scheme writing similar to BNF schemes.

View more solutions

Comments

[lavinio]

You can put XSD inline with XML. Just use the right namespace, and nest the schema inside the document. WSDL files do this commonly. DTDs are subject to DoS attacks. See en.wikipedia.org/wiki/Billion_laughs

[Yinda Yin]

@lavinio, I think you're right about inlining XSD; I have seen files before that do this. Interesting article about the Billion Laughs attack.

[dma_k]

Is there any DTD feature, that cannot be done in XSD?

[Scott Chu]

Yes! EditiX provides a way to write in Relax NG and convert it to DTD or XSD. I pro the Relax NG. XSD is sometimes kinda complicated for some of my junior engineers.