Tim Bray Says RELAX

twofish writes to tell us that Sun's Tim Bray (co-editor of XML and the XML namespace specifications) has posted a blog entry suggesting RELAX NG be used instead of the W3C XML Schema. From the blog: "W3C XML Schemas (XSD) suck. They are hard to read, hard to write, hard to understand, have interoperability problems, and are unable to describe lots of things you want to do all the time in XML. Schemas based on Relax NG, also known as ISO Standard 19757, are easy to write, easy to read, are backed by a rigorous formalism for interoperability, and can describe immensely more different XML constructs."

Comments

Don't do it.

[Anonymous Coward]

When you want to come.

Couldn't agree more

[antonyb]

My experience with XML Schema is exactly that; hard to write in the first place, hard to maintain, and regular interop problems between different implementations that make the theory of web services a practical nightmare (idrefs are the first example that spring to mind).

On the other hand, RELAX NG "just works".

(all IME of course...:)

ant.

I have to agree.

[JanusFury]

Has anyone here ever tried to read an XML schema for anything relatively complex? It's a nightmare. RELAX looks much cleaner and more direct, which I wholeheartedly approve of.

Re:I have to agree.

[sien]

Yes. I've done it using Relax NG and it was easy, simple and readable.

It also works really, really well with the nXML mode for emacs.

Finally, XML schemas in a way that are not verbose, ugly and unreadable. And if you do need one of the older schema languages there are translators from RelaxNG available.

Re:I have to agree.

[radtea]

I was at SGML '96 where XML was first announced, and was one of those people who went home and wrote a (non-validating) XML parser over the weekend, based on the draft spec. I've used both DTDs and XML Schemas and can say without question that schemas are actually a bigger pain to work with than DTDs. DTDs were bad enough, but schemas have been a major step backwards, adding complexity without adding the features one actually needs.

Some years ago I wrote a code generator that used DTDs as the data modelling language. I sold it to the company I was working for at the time and someone I had no control over re-wrote it use schemas because they were "simpler". The result had major bugs and dropped features, not entirely due to schema-related problems, although it is worth noting that the "simplifications" included handling schemas in completely incorrect ways, because if you handled them correctly they could not do the job. I created a new generator from scratch last year and tried to do thing "properly" with schemas. It was essentially impossible, and I wound up creating a custom XML-based language use as input.

At the time there was no Relax NG standards process, so I stayed clear of it. But it has the blessing of James Clarke too (author of the SP SGML parser and the expat XML parser.) So it is probably worth another very hard look.

Re:Just sit back...

[ubernostrum]

What kind of programmer can't use XML effectively anyhow...oh wait... (No, I didn't read TFA!)

Helpful hint for understanding the above: Tim Bray, author of TFA, is one of the guys who originally developed and spec'd out XML. Really. His name's on the spec and everything. So if he says that a particular XML tool has problems, it's probably a good idea to take him at his word ;)

Re:it's a rather straightforward observation

[Peter Cooper]

Check out YAML.

Maximizing Composability and Relax NG Trivia

[SimHacker]

Tim Bray is right, and he couldn't have put it better: W3C XML Schemas (XSD) suck. The reason Relax NG is so much cleaner and more powerful than committee-designed XML Schemas, is that it's based on a sound mathematical foundation (tree regular expressions, or "hedge automata theory"). While XML-Schemas suffer from ad-hoc design, committee-burn, lack of focus, and half-baked attempts to solve too many unrelated problems.

Here's some interesting stuff from my blog about the design and development of Relax NG.

-Don

James Clark wrote about maximizing composability:

First, a little digression. In general, I have made it a design principle in TREX to maximize "composability". It's a little bit hard to describe. The idea is that a language provides a number of different kinds of atomic thing, and a number different ways to compose new things out of other things. Maximizing composability means minimizing restrictions on which ways to compose things can be applied to which kinds of thing. Maximizing composability tends to improve the ratio between functionality on the one hand and simplicity/ease of use/ease of learning on the other.

Clark describes the derivative algorithm's lazy approach to automaton construction:

I don't agree that <interleave> makes automation-based implementations impossible; it just means you have to construct automatons lazily. (In fact, you can view the "derivative"-based approach in JTREX as lazily constructing a kind of automaton where states are represented by a canonical representative of the patterns that match the remaining input.)

The Relax NG derivative algorithm is implemented in a few hundred elegent declarative functional lines of Haskel, and also in tens of thousands of lines and hundreds of classes of highly abstract complex Java code.

Clark's Java implementation of Relax NG is called "jing", which is a Thai word meaning truthful, real, serious, no-nonsense, and ending with "ng".

Comparing the Java and Haskell implementations of Relax NG illustrates what a wicked cool and powerful language Haskell really is. The Java code must explicitly model and simulate many Haskel features like first order functions, memoization, pattern matching, partial evaluation, lazy evaluation, declarative programming, and functional programming. That requires many abstract interfaces,, concrete classes and brittle lines of code.

While the Java code is quite brittle and verbose, the Haskell code is extremely flexible and concise. Haskell is an excellent design language, a vehicle for exploring complex problem spaces, designing and testing ingenious solutions, performing practical experiments, weighin

Re:XML Totally Sucks - All of it!

[Just Some Guy]

While XML may have it's places (I've yet to encounter one in the commerical world), passing large amount of data is not one of them.

Yeah, well I have to look at EDI every day. I'd switch to XML in a heartbeat if it were up to me.

You picked some obvious strawmen to shoot down. XML isn't for building gigabyte databases (regardless of whether some people try to use it for that). It's for easily moving data between applications. If you think writing a flat text parser is easy, then you've never had to deal with nested data or escaped characters. Say what you will about XML, but it's nice to have one set standard that deals with all that, even if suboptimally, because I never want to write another ad-hoc parser for as long as I live. Been there, done that, have no desire to bother again.

XML uses a binary format

[ClosedSource]

Of course ASCII (or UNICODE for that matter) is a binary standard as well. So special tools called text editors were created so that people could read it.

There are more sophisticated binary standards that are more efficient than ASCII and it wouldn't take a lot of effort to create viewers/editors for them as well. Of course most markup documents would be significantly smaller if tags didn't have to be S-P-E-L-L-E-D O-U-T character by character. Each HTML tag could be encoded in just two bytes with lots of room to spare.

It always fascinates me that we have no problem making customers use a new specialized tool like a browser, but it's taboo to use a non-ASCII tool for development. So we continue to structure our data as if it were going to be processed by a VT100.

Re:One fix to XML I'd like to have...

[nuzak]

That feature is in SGML. In fact it can be even shorter than that, you can express an entire tag and its content with is optional). SGML even lets you change the angle brackets to anything else you want. You can make any SGML doc look like nothing you or anyone else has ever seen ... all part of the feature set.

SGML is full of fun little hacks like that, and it was a pain in the ass to read. Omitting the tag name from the end tag makes it impossible to know you have an improperly closed tag til the end of the document, and then you have no idea which tag wasn't closed. And no, that wasn't a theoretical problem either, this became a real problem with giant SGML docs that used all the shortcuts.

If you really hate XML's verbosity so much, realize that it was designed for easy reading, not easy writing. I whipped up my own xml mode in emacs and made '</' trigger an "electric-slash" behavior that closes the tag automatically. Not rocket science.

XML nightmare

[rgaginol]

If XML Schema was a work colleague they would be Wally from Dilbert - it's not that things are impossible to do with it, it's just that the relative simple things become hard and the complex almost impossible. Due to the fact that almost anything is possible with XML schema with enough work (weeks, months years...) instead of just scrapping it, people keep at it doggedly despite the number of times we get bitten. I'd love to see the community move more completely to RELAX NG if it makes my life easier.

XML is like Electricity

[SimHacker]

It's good for transmitting information/energy, but it's not good for storing it.

-Don

I call this the LineOfView (as in PoV) Problem

[Qbertino]

I call this the Line of View (as in PoV) or 'Horizon' Problem. The general problem is this: In XML we've got a standard that is universal for displaying n-dimensional structures in a basically 1-dimensional enviroment. (For the time being, we're ignoring that XML text ususally goes from left to right and top to bottom, making that something 2D to look at)
The question now is: where do you draw the line of view? Along which line do I take my knife to cut open my n-dimensional structure to unravel it and flatten it out into a 1-dimesional string of characters? This is a problem that is impossible to solve satisfactory for all possible PoVs or - as I say - Lines of View, or better yet, Horizons to the structure. Will I unravel my DB of books by authors? By issues? By vendors? By publishers or by weight and size? ... At some point you will have to look at in which way you want to handle your stuff and which way you're going to unravel it. This will undoubtly influence on how much XML clutter you will have to construct. With XML it's the same as with databases: It/they will allways be pathetic crutches for us to latch on to the real work. Undispensable, but crutches nontheless.

What I'm getting to is this: mapping n-dimensional stuff to 1-dimensional structures will allways suck one way or the other. It's just that with XML we all start agreeing upon in which way it's supposed to suck. I don't think that changing the Schema standard (or worse: introducing additional standards) will actually attack this hard problem. I have a strong suspicion that Relax NGs relief is illusional, short term and re-introduces downsides that XML Schema allready has takled with it's pesky and strict nature. For one it would be consistency with the View-Horizon once chosen all the way through the given data-structure. I don't know for shure - go test and find out - but I do know that universal serialization will allways come with downsides and RelaxNG (or any other schema) won't change that.