The havoc the XML purists wrought | Hacker News Reader

The havoc the XML purists wrought

scripting.com

31 pointswhyleyc33 comments

They say that XML is too complicated, but that's wrong. Just ignore everything but elements, attributes and namespaces. That's what RSS does, and it's very simple. [..] I don't trust would-be platform vendors who won't accomodate all possible developers, esp those who use a language that's so deeply installed as XML is.

He talks as if JSON is somehow "hard" for developers to grok. It's about as simple as you can get. I don't think it's noble to support a subset of XML, where regular JSON would fit better, merely to placate some developers used to doing things the hard way. If anything, API developers should be leading us by the nose to do things the best way.

Even if you "ignore" much of XML, as Winer suggests, you can still stumble into trouble. I've written (and used) a few RSS parsers in my time and dealing with broken XML with a regular XML parser is a gigantic pain in the ass. Many developers don't sanitize their input properly or dump source HTML or XHTML verbatim into their <description> elements. Try throwing that through Expat without thinking about it.. You can screw up JSON too, of course, but it's almost entirely down to quote marks alone.. not tagging, erroneous character encodings, non-existence namespaces, and more.

In contrast, Amazon supports XML in their web services, quietly and competently. I can't imagine them saying one day "It's too much work for us to keep supporting XML so you all have to rewrite your code now if you want to keep paying us for the web services you use."

Amazon's Web Services are good/unique enough for people to put up with the bullshit of XML in order to use it.. though their command line and Web interface tools are good enough that I suspect most users never need to get down to dealing with XML anyway.

There are plenty of complicated things that aren't complicated anymore if you ignore the complicated parts. (For example, the global financial system.) I don't think that line is a valid argument.

XML is more complex than it needs to be, but this article is just ridiculous. XMPP isn't complicated because it uses XML, it's complicated because it's a modern IM protocol. If anything, using XML makes implementing an XMPP library significantly easier than if it had its own wire format.

A better comparison than RSS v. XMPP is RSS v. Atom. And in this comparison, RSS is by far the loser. It's incredibly difficult to write an RSS parser which can handle even a small fraction of the RSS published today, mostly because the standard is absolute garbage. In contrast, an Atom parser can be knocked out in an afternoon.

I'd be happy with a "reduced set" of XML, which excludes stuff like DTDs, named entities, and references. I've never seen these features put to any significant use in real life. But most of XML is quite sane (if a bit verbose).

If we follow Winer's advice and use just a simple subset of XML, what does XML buy us that JSON doesn't also provide with less overhead and better readability? (Yes, I know JSON doesn't have namespaces, but I'm not persuaded that namespaces belong in any subset of a data format that can be called "simple".)

XML Schema is hilariously painful but it at least exists and is in fairly common use. JSON Schema is still an Internet-Draft that almost nobody seems to be aware of, so JSON-based systems are still at the "garbage in, garbage out" level and will only get worse with maintenance.

http://tools.ietf.org/html/draft-zyp-json-schema-02

XML Schemas are well established in areas like insurance. I worked on a insurance portal in 2001-2002 that made extensive use of the Origo schemas - I realized that XML is really huge in some areas. To the non-enterprise / HN startup / web dev types it seems that XML can be wished away, but they have no idea of how well entrenched it is.

When I've designed XML files in the past, I use the "tags and attributes" principle. The entire file is tags, and ignored whitespace between the tags. All the data is in attributes. Looks a lot nicer IMO.

What I don't get, with XML in the real world, is why people prefer this...

instead of my preference...

With the first form, you need to write out each tag name twice, and there's ambiguity when whitespace appears between tags. Yet that's what most XML out there looks like. Maybe I'm just odd,

A choice between a node and an attribute is not always clear, but there are helpful guidelines out there. A search for xml nodes versus attributes will give you a list of articles to ponder -- sorry for not being more detailed, but I am away from my archives, or else I could have given you links to the places that helped me to visualize it better. Here is one classic, though, from as far back as 1992: http://xml.coverpages.org/attrSperberg92.html

EDIT: ok, the first link is to an IBM Research article which I recall having been useful: http://www.ibm.com/developerworks/xml/library/x-eleatt.html

There's a semantic difference between an element and an attribute. Unfortunately, XML is specifically used to model element-as-attribute relationships, so a lot of people start wondering about the exact, nebulous, distinction. Elements are the more flexible of the two, but possessing of some quality that you (and many others) wish to avoid.

The difference is much the same as that between native datatypes and objects. Some languages explicitly expose everything as objects, while others contend that, for example, boolean-objects are overkill and downright problematic.

Generally, I use attributes for strongly typed data, and elements for untyped or (semi)structured data. Schemas are vital for typing.

I always use the same example as why I dislike XML so much

    <a>
        <b>hi</b>
    </a>

is not the same as

    <a><b>hi</b></a>

But Erik Naggum makes the argument against XML so much more fun

http://harmful.cat-v.org/software/xml/s-exp_vs_XML

In what contexts would those two snippets of XML be treated differently? (Not that I am jumping in to defend XML I'm just curious).

If you wrote code to crawl the tree generated from that XML, you'd have text nodes (of whitespace) in one, but not the other.

That's what I thought - I don't think I've ever seen an application that would bother with that difference.

You're using that application right now - XHTML takes the difference into consideration. Consider the <pre /> tag:

  <pre><b>Hello!</b></pre>

is rendered differently from

  <pre>
     <b>Hello!</b>
  </pre>

Good point! :-)

The difference is a big deal to applications that use XML as a markup language (marking up documents and text), as it was designed to do originally.

How are those two examples different? Every XML system I've used would see those as the same.

I, however, really enjoyed this line from the link. "I once believed that it would be very beneficial for our long-term information needs to adorn the text with as much meta-information as possible. I still believe that the world would be far better off if it had evolved standardized syntactic notations for time, location, proper names, language, etc, and that even prose text would be written in such a way that precision in these matters would not be sacrificed, but most people are so obsessively concerned with their immediate personal needs that anything that could be beneficial on a much larger scale have no chance of surviving."

This wonderfully, succinctly explains why efforts like the semantic web are doomed to failure.

Technically, they are different. In the first example, there's whitespace between <a> and <b> as well </b> and </a>. Most applications ignore whitespace but they're not required to - whitespace is not ignored in XML.

I just died a little inside. Knowing that there are people out there that would inflict this kind of nonsense as a standard for others to use makes one really lose faith in humanity.

That's the only way to distinguish

  an <strong>emphasized</strong> word

from

  an<strong>emphasized</strong>word

Which makes perfect sense for documents. So why do we use a markup language suitable for documents for general data? Whitespace matters in a document, but does not matter for data. (I'm not accusing you, I'm actually curious if you know the answer. I've never had to deal with XML.)

It may matter a lot for some data... When you send a binary file, would you like all of the zeros stripped? Your insignificant data from one perspective is possiblely significant from another perspective.

I don't follow. If you need literal, unstructured data embedded in a structured format, then use whatever escape mechanism the format has.

It is not possible to ship binary over XML. Regardless of encoding used, XML rigidly forbids the bytes corresponding to non-whitespace ASCII chars below 32: http://www.w3.org/TR/2008/REC-xml-20081126/#charsets This makes arbitrary binary impossible. You have to base64 encode it or something.

Not entirely relevant to your point, but worth bringing up. (XML doesn't ever mangle binaries, because it simply forbids them.)

Because for the longest time there was nothing both better and standardized. I've dealt with xml many times and hate it but It's better than maintaining your own proprietary DTL.

Anyone who adopted JSON could have been using s-expressions all along, but it's become a tradition to ignore everything Lisp got right until some brave soul gives it a makeover with curly braces.

In theory, I'll agree with you. However, as far as I know, there are no s-exp parsing libraries for any language I know but Lisp. Also s-exps as they exist in Lisp assume some Lispish concepts that don't map to all languages. Symbols, for example, don't exist in Java. JSON is a better solution for cross language portability.

(sorry I want to reply to the reply but there's no reply button for it)

Saying there are no s-expression parsers for non-Lisp is ridiculous.

This one's http://sexpr.sourceforge.net/ for C, it's been in sourceforge since 2002.

perhaps you prefer one in :

Javascript - http://planet.plt-scheme.org/package-source/dherman/javascri...

Perl - http://search.cpan.org/~nelhage/Data-SExpression-0.34/lib/Da...

Limbo - http://man.cat-v.org/inferno/6/sexprs

I shan't go on

I didn't say that there weren't s-expression parsers for languages other than Lisp; just that I haven't found them. I'll be the first to admit that I haven't really been looking.

But you'll note the disparity in support. A library for parsing XML with C may have existed since 2002, but it's existed for XML since 97. In Java, XML support has been baked into the JVM for a while, but S-exp parsers are hard to find. That library you list for Javascript is not a s-exp parser but rather a Scheme library for generating Javascript from a Scheme-like language. Googling for javascript and s-exps returns a lot of results like that but not many s-expression parsers. There certainly aren't any widely used and recognized libraries like there are for XML and JSON fro most languages aside from Lisp.

It's an excellent question. My answer is, use XML for data if you need features that come with it, such as validation, or if you'll want to use it with tools/standards that orbit XML, like XQuery, or if you have a situation where you're dealing with both documents and data, and you want them to interoperate. Otherwise there's probably not a lot of benefit to XML; JSON, YAML, or a DSL will suit you just fine.

A lot of XML-hate comes from its overuse—it comes with overhead that is too expensive for simple cases, but is pretty useful when you go beyond simple. Right tool for the right job and all that...

The line breaks are classed as nodes (as is all whitepsace) unless you tell XML parsers to ignore them usually.

I've become a JSON and YAML fan as I think they have higher signal to noise ratios (they each have their own issues with white space though).

  JSON:
  {"a":{"b":"hi"}}

  YAML:
  ---
  a:
    b: hi

You're complaining that standard parsers give you extra information about the document being parsed? Another thing that might annoy you: even though comments are "ignored", many XML parsers generate comment nodes in the DOM. This allows you to produce an output document that is exactly equivalent to the input document.

If a node doesn't mean anything to your application, ignore it. If you're an XHTML parser, the whitespace is significant, so you need to handle it. If you're a FooML parser, the whitespace is insignificant, so just ignore it.

If you hate XML, this is not a particularly good justification.

Reply on news.ycombinator.com