News for hackers. Stuff that matters.

A comment on Using Standard HTML (1997).

Re:Using Standard HTML (1997) by nayuki on Sunday January 19, 2020 @09:52AM (#22091693)
Whenever I write HTML documents, I find it helpful to use the XHTML syntax. The strict XML parser catches syntax errors, incorrect nesting, unescaped characters, unquoted attributes, and more. And it supports all the tags and attributes of the HTML5 living standard, so I don't miss any modern functionality.

The standard "tag soup" HTML parser has numerous surprising behaviors on invalid or questionable HTML code. Examples:

The <p> element cannot nest, so starting a new one will end the previous one:

  <p>Hello
  <p>World
  
  gets transformed to
  
  <p>Hello</p>
  <p>World</p>
Furthermore, trailing junk triggers weirdness:

  <p>Hello
  <p>World
  </p></p>
  
  becomes
  
  <p>Hello</p>
  <p>World</p>
  <p></p>
Some elements are defined in the HTML DTD to have implicit start tags:

  <table><td>abc</td></table>
  
  yields the DOM tree of
  
  <table><tbody><tr><td>abc</td></tr></tbody></table>
Whereas in XHTML mode, either your code is a parse error, or what you see is what you get. You can nest <p> elements in XHTML, and you can make a <table> contain <td> directly (without the <tbody> and <tr> containers).
[ Parent | on HN ]
Re:Using Standard HTML (1997) by tannhaeuser on Sunday January 19, 2020 @12:41PM (#22092665)
In case you haven't already, I think you should take a look at SGML ([1], my site) which standardizes these and all other tag omission/inference rules used by HTML, and also attribute minimization. The hard-coded tag soup parser is by no means "standard".

[1]: http://sgmljs.net

[ Parent | on HN ]
Re:Using Standard HTML (1997) by rimliu on Sunday January 19, 2020 @03:40PM (#22093901) 

  > The standard "tag soup" HTML parser has numerous
  > surprising behaviors on invalid or questionable HTML code.
There is nothing surprising here. And some elements are allowed to omit end tags, or even start and end tags altogether.

In fact, if you dig deeper there are a lot more surprises in XHTML documents (assuming they are served with the proper MIME type so they get treated like XHTML).

[ Parent | on HN ]
Re:Using Standard HTML (1997) by JoshTriplett on Sunday January 19, 2020 @02:28PM (#22093390)
Unfortunately, the XML-style syntax with all tags closed isn't allowed in HTML for certain tags. For instance, you can't close <input> or <br> or <hr> tags.
[ Parent | on HN ]
Re:Using Standard HTML (1997) by firefoxd on Sunday January 19, 2020 @02:34PM (#22093431)
Yes you can.

    <input />, <br />, and <hr />
And it is valid XML. They are referred to as self closing tags.
[ Parent | on HN ]
Re:Using Standard HTML (1997) by rimliu on Sunday January 19, 2020 @03:39PM (#22093890)
It is valid XML. It means totally different thing in HTML.
[ Parent | on HN ]
Re:Using Standard HTML (1997) by detaro on Sunday January 19, 2020 @03:44PM (#22093923)
The self-closing form is fine in HTML5 for those examples, since they are void elements. The "/" gets ignored, which is what you want in that case.

It fails if you try this with non-void elements: E.g. XHTML <div /> might turn into <div> in HTML5, which opens a div element, which is not the same. (I'm not sure if this behaviour is specified or not)

https://html.spec.whatwg.org/multipage/syntax.html#start-tag...

> 6. Then, if the element is one of the void elements [...] then there may be a single U+002F SOLIDUS character (/). This character has no effect on void elements [...]

https://html.spec.whatwg.org/multipage/syntax.html#void-elem...

> area, base, br, col, embed, hr, img, input, link, meta, param, source, track, wbr

[ Parent | on HN ]
Re:Using Standard HTML (1997) by vbezhenar on Sunday January 19, 2020 @10:23AM (#22091866)
Does every popular browser support it in the same way? I would expect some subtle differences, because this mode is not very popular and probably not as tested as standard HTML.
[ Parent | on HN ]
Re:Using Standard HTML (1997) by Terretta on Sunday January 19, 2020 @10:40AM (#22091970)
On the contrary, part of the point of XHTML is you can reference the document type definition (DTD) so everyone would be on the same page.

https://www.w3.org/wiki/Doctypes_and_markup_styles

For example...

    <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
... is the XHTML 1.0 equivalent of HTML 4.01 except it won't validate loose HTML style markup, it all needs to be well formed XML.

That said, for HTML5 ...

https://stackoverflow.com/questions/1570418/if-i-use-html-5s...

Ten years ago they thought that should work a while. Looks like it.

[ Parent | on HN ]
Re:Using Standard HTML (1997) by vbezhenar on Monday January 20, 2020 @05:06AM (#22097772)
Thanks, those are definitely good things. I never had much thought about it, but I definitely was bitten more than once by those auto-closing and auto-generating HTML tags, so strict XML might be a good alternative. Definitely will try it next time.
[ Parent | on HN ]
Re:Using Standard HTML (1997) by garaetjjte on Monday January 20, 2020 @11:12AM (#22099806)
Note that XHTML must be served with application/xhtml+xml content-type.
[ Parent | on HN ]
Re:Using Standard HTML (1997) by nayuki on Sunday January 19, 2020 @04:13PM (#22094118)
Gonna aggregate all the subtopics in one reply.

> you should take a look at SGML

I am aware that SGML is the ancestor of HTML and XML. I'm not sure there are any popular software today that handle SGML syntax. Furthermore XML has fewer rules to remember than SGML. So I will largely ignore SGML, except for explaining why some behaviors went into HTML.

> The hard-coded tag soup parser is by no means "standard".

I believe HTML5 standardized the syntax/error handling for HTML (not XHTML) documents. It should mean that for any given document, every HTML5-compliant web browser - Chrome, Firefox, Edge, etc. - will parse it into the same DOM tree.

> Does every popular browser support [XHTML] in the same way?

Yes, XHTML mode is supported by IE 9+, all Firefox, all Chrome, all Safari, all Opera, and all mobile browsers on Android and iOS. https://caniuse.com/#feat=xhtml

> I would expect some subtle differences, because this mode is not very popular and probably not as tested as standard HTML.

Indeed XHTML is very unpopular. There was one slight bug I encountered in Firefox about a year ago, where invalid XHTML pages would show a yellow error screen (correct) but with no error message text (this was a regression). That said, I'm pretty sure that the XML-parsing engine, XSLT, and related technologies are here to stay, so XHTML isn't at risk of losing support.

> On the contrary, part of the point of XHTML is you can reference the document type definition (DTD) so everyone would be on the same page.

That was the original intent in XHTML 1.0, 1.1, and 2.0. But anyway, nowadays the DTD is optional and probably unhelpful: https://stackoverflow.com/questions/38427011/does-xhtml5-req...

>> The standard "tag soup" HTML parser has numerous surprising behaviors

> There is nothing surprising here.

So are you an expert at diagnosing HTML parsing and invalid syntax? I prefer not to be one, because I'd like the XML parser to yell at me loudly when I'm wrong, rather than helpfully try to fix my code the wrong way.

> if you dig deeper there are a lot more surprises in XHTML documents

Like what? I'll try to name a few: Namespaces (colons), character entities, CDATA sections, strict comments (rules about the substring "--").

> assuming they are served with the proper MIME type

Yes, I forgot to mention this. My website https://www.nayuki.io/ has been serving application/xhtml+xml for about a decade, and seems to work well with the browsers that the audience uses.

[ Parent | on HN ]