Talk:XML Schema Language Comparison
From Wikipedia, the free encyclopedia
Contents |
[edit] Initial Creation
This page was created from a section in the RelaxNG page, detailing the differences between RelaxNG and W3C XML Schema. Korval 20:24, 4 December 2006 (UTC)
[edit] NPOV suggestion
The section of claimed advantages of XSD over RNG is not exactly adhering to the NPOV. All the advantages claimed could be considered misfeatures of XSD and, therefore, the lack of those features in RNG could be considered RNG's advantages over XSD. Hsivonen 09:13, 19 December 2006 (UTC)
- I disagree. While it is possible to consider some of WXS's features to be misfeatures, that should only apply where there is a good faith reason for it. I'll take the advantages one by one:
- PSVI, for example, is a reasonable feature, even if one might think that such shouldn't be applied to a schema language. It has uses, and these uses are objectively useful. Where PSVI breaks down is that, to make it happen, some of WXS's potential specificity had to be sacrificed. Since the RelaxNG section speaks to RelaxNG's greater ability to define patterns, there is no need to point out that PSVI is one of the reasons for the disparity in specificity between the two.
- As another example, it was cited on the original article in RelaxNG that RelaxNG's inability to specify specific lengths of sequences could be considered a feature. I fail to see the significance of it, for two reasons. One, the Zero-One-or-Infinity "rule" is, while a nice thing in theory, isn't always wise in practice. Two, you can still write WXS schema's under the ZOI rule; just don't use that greater specificity. RelaxNG schemas must operate under ZOI, whether it is reasonable or not.
- I specifically mentioned that the WXS ability to specify default data was a double-edged sword in the disadvantages of WXS section. I saw no need to repeat that in the RelaxNG section (which should be about what RelaxNG does, rather than what it does not do).
- As far as data types are concerned, I mentioned that most RelaxNG implementations support the standard WXS-defined types. And as for having a way to bind a WXS to an XML file... I fail to see any negative consequences of that. It is still up to the parser to deal with it.
- So, in conclusion, I would say that if anyone isn't adhering to NPOV, it's probably more you than the article. I have a strong dislike for WXS, but I did give it reasonable consideration in this article. You may want to verify that some of what you consider misfeatures comes from whether they are misfeatures or what you personally think of these features.Korval 10:49, 21 December 2006 (UTC)
-
-
- I have some concerns with most of your edits, and I'll be modifying them appropriately. First of all, most of your additions were already mentioned in the WXS section under its disadvantages. As an example, you mention the dangers of default data; this was already in the WXS disadvantages section. The commentary on data type extensibility was likewise mentioned as both an advantage of RNG and a disadvantage of WXS.
-
-
-
- While your section on the Zero/One/Infinity rule is not repetitious, I see it as a false negative. I'm not a big fan of rules of thumb, and ZOI isn't one of my favorites. I understand where it comes from, but I prefer flexibility over hard-and-fast rules. So I'm finding it rather difficult to see how an explicit restriction is an advantage. It's no different from how RNG's ability to combine elements and attributes arbitrarily is more powerful than WXS's, even though WXS had specific reasons for limiting this construct.
-
-
-
- I'm still not clear on what the problem with PSVI is. I know what the problem with infoset augmentation with default arguments is; I mentioned it earlier. The problem cited in Tim's post with PSVI doesn't seem to be specific. He basically just says, "there's nothing wrong with a Type-Augmented Infoset (TAI), but why link it to validation?" Effectively, he's saying that coupling infoset building with validation is wrong, but there is no justification as to why it is wrong. Simply saying that putting two things together wrong does not make it wrong.
-
-
-
- The other arguments mentioned on Tim's thread are largely meaningless today. XQuery, XPath 2.0, and XSLT 2.0 are not defined in terms of PSVI; they use the standard XML infoset. PSVI's intended use is clearly conversion between XML and objects in a programming language; it has not spread to "infect" the rest of XML's core. Korval (talk) 02:31, 10 December 2007 (UTC)
-
[edit] Timeline Talk
Please address which schema languages are the most recent. It is quite confusing to read about XML schema validation methods, tools, frameworks, etc since they usually are stuck with whatever schema validation language existed at their initial development. This can lead one to select an obsolete or soon to be obsolete technology (e.g., DTD) for XML validation.
- I agree with the suggestion for some discussion about tool availability. I'm not sure how the timeline with regard to creating each of these schema languages matters.Korval 00:30, 26 January 2007 (UTC)
[edit] CFeet77's changes
Your changes removed information about RELAX NG's compact syntax. If you wished to consider the greater specificity to be more important, then remove the statement about the importance of the compact syntax. Furthermore, everything you discussed was mentioned, in far greater detail, below your comment. As such, it was entirely redundant and unnecessary.Korval 19:53, 26 January 2007 (UTC)
