Tag soup

In Web development, tag soup refers to formatted markup written for a Web page which is very much like HTML, but may not consist of correct HTML syntax and document structure. Because web browsers have historically treated HTML syntax or structural errors leniently, there has been little pressure for web developers to follow published standards, and therefore it is necessary that all browser implementations treat what looks like HTML as "tag soup", accepting and correcting for invalid syntax and structure.

An HTML parser, such as a web browser, which is capable of interpreting HTML-like markup even if it contains invalid syntax or structure may be called a tag soup parser. All major web browsers currently use a tag soup parser when interpreting HTML or content it believes to be HTML.

Tag soup may collectively refer to a large number of common authoring mistakes, such as malformed HTML tags, improperly-nested HTML elements, and unescaped character entities (especially ampersands (&amp;) and less-than signs (&lt;)).

I have used this term in my instruction for years to characterize the jumble of angle brackets acting like tags in HTML in pages that are accepted by browsers. Improper minimization, overlapping constructs ... stuff that looks like SGML markup but the creator didn't know or respect SGML rules for the HTML vocabulary. In effect a soupy collection of text and markup. [...] I've never seen the term defined anywhere.

Alternatively, the term tag soup may occasionally be applied to any markup which uses poor methodology such as the misuse of HTML elements for semantic purposes other than which they were intended. Though not strictly syntactically or structurally incorrect, such code can be hard for a human reading the source code or a non-visual user agent to interpret meaningfully.

Overview
Tag soup is a term used to denigrate various practices in web authoring. Some of these (roughly ordered from most severe to least severe) include:


 * 1) Malformed markup where tags are improperly nested. For example, the following:
 * 2) Invalid structure where elements are improperly nested according to the DTD for the document. Examples of this include nesting a "ul" element directly inside another "ul" element for any of the HTML 4.01 or XHTML DTDs.
 * 3) Use of proprietary or undefined elements and attributes instead of W3C defined ones.
 * 4) Use of non-SGML comment syntax in HTML versions up to 4.01.  HTML versions up to 4.01 are supposed to be based on SGML, but the SGML comment syntax is rarely used and not supported by most browsers, thus such non-SGML comments may be considered "tag soup".  For example:
 * 5) Use of XHTML syntax in a document which is served as HTML.  Serving XHTML code to browsers and telling them it is HTML constitutes serving tag soup, because the browser has no choice but to interpret it as HTML-like markup with "syntax errors".  This particular use of tag soup is recommended by Appendix C of the XHTML 1.0 specification as a way of serving XHTML to browsers that do not support it.

Malformed markup
Malformed markup is arguably the most severe problem in web authoring. However, thanks to better education and information and perhaps a butterfly effect from XHTML, the issue of malformed markup is becoming less common. Browsers, when faced with malformed markup, must guess the intended meaning of the author. They must infer closing tags where they expect them and then infer opening tags to match other closing-tags. The interpretation can vary markedly from one browser to the next. Ian Hickson wrote a detailed article investigating the differences between how browsers handle tag soup.

While many graphical web editors produce well-formed markup, an author writing code manually with a text-editor and then testing only in one browser can easily miss such errors. The presentation can therefore vary drastically from one browser to another as each tries to “correct” the authorʼs intent in different ways and then applies styling to those “corrections”.

Invalid document structure
Invalid document structure here means only the use of attributes and elements where they do not belong. For example, placing a "cite" attribute on a "cite" element is invalid since the HTML and XHTML DTDs do not ascribe any meaning to that attribute on that element. Similarly, including a "p" element within the content of an "em" element is also invalid. With the move toward separating malformed markup from invalid markup, the problems with invalid markup have increasingly been seen as less severe. Some have begun to advocate looser content models that allow greater flexibility in authoring HTML documents (whether in HTML or XHTML). However, use of invalid markup can blur the author's intended meaning, though not as severely as malformed markup.

Many graphic web editors still produce invalid markup. Moreover, many professional web designers and authors pay little attention to issues of validity. It is common to see invalid markup in many of the sites throughout the World Wide Web.

Use of proprietary/discontinued elements
In the early age of the web (much of the 1990s), the design of the official HTML specification became increasingly strained, compared to the desire of designers for flexibility in creating visually vibrant designs. In response to this pressure, browser makers unilaterally added new proprietary features to HTML that fell outside of the standards at the time. This meant there were proprietary elements in HTML that worked in some browsers, but not in others.

To some extent, this problem was slowed by the introduction of new standards by the W3C, such as CSS, introduced in 1998, which helped to provide greater flexibility in the presentation and layout of web pages without the need for large numbers of additional HTML elements and attributes.

In later standards, many elements have either been combined into a single semantic construct (such as object elements replacing proprietary applet, and embed elements) or have been deprecated (such as the "s", "strike" and "u" elements). Nevertheless, browser developers have continued to introduce new elements to HTML when they have perceived a need. Some browsers include tabindex attributes on any element. WebKit developers aligned with Apple introduced the "canvas" element that behaves much like the "object" or "embed" element. Mozilla then introduced their own "canvas" element, which behaves even more like the "object" element.

HTML compared to XHTML
With the exception of "HTML-compatible" XHTML 1.0 as defined in XHTML 1.0 Recommendation Appendix C, XHTML requires browsers to avoid rendering a page at all if it contains any invalid syntax. This has led some to refer to all forms of HTML as tag soup. Web browsers shoulder the greatest burden in parsing and simultaneously correcting malformed "HTML" markup. In contrast, web browsers need to perform no corrections whatsoever on syntactically invalid "XHTML" markup. The browser is expected to simply fail in rendering. An error message is displayed pointing the user to the first instance of malformed content. XML parsing is greatly simplified by this practice compared to "HTML" parsing. All "HTML" parsers might be called tag soup parsers: meaning everything interpreted as an HTML-like document is parsed as tag soup.

Evolving specifications to solve tag soup
While some of the issues of tag soup are due to shortcomings of browsers and sometimes due to a lack of information for web authors, some of the proliferation of tag soup was due to missing links in the web standards themselves. The W3C has spearheaded several efforts to address the shortcomings of web standards. As more browsers support newer revisions of standards, the pressure on web developers to use non-standard code to solve problems diminishes.

Cascading Style Sheets (CSS)
Cascading Style Sheets (CSS) provide a mechanism to specify the presentation of elements in a document without altering the markup structure of the document. Before CSS was commonplace, web developers may have resorted to some structurally invalid markup to achieve certain presentational goals - for example, including block level elements within inline elements to obtain a particular effect.

XML and XHTML
XHTML is a reformulation of the HTML language based on XML. XHTML was developed to address many of the problems associated with tag soup.

First, XML allows parsers to separate the process of interpreting the document syntax and its structure. In HTML and SGML, a parser needed to know certain rules about elements during parsing, such as what elements could be contained within other elements and which elements implicitly close the previous element. This is because in HTML and SGML, closing tags and even opening tags were optional on some elements. By requiring all elements to have explicit opening and closing tags, XML parsers can parse the document and produce a document tree without any knowledge of the document type. This allows parsers to be universal and very light-weight, and to be separated from the process of validating or interpreting the document.

Second, the XML Specification clearly defines that a conforming user agent (such as a web browser) must not accept, and not continue parsing, if any syntactical error is encountered. Thus, a browser interpreting a Web page as XHTML will refuse to display the page if it encounters a formation error. This can help ensure that when authors test XHTML code against a conforming browser they will immediately be informed of malformation problems: perhaps the most severe problem facing web browsers. When code is malformed, the intent of the author is ambiguous. Without the directives of XML, HTML browsers must have specific algorithms to infer the author's intended meaning in a wide range of cases where invalid syntax is encountered.

Third, XML and XHTML introduce the concept of namespaces. With namespaces, authors or communities of authors can define new elements and attributes with new semantics, and intermix those within their XHTML documents. Namespaces ensure that element names from the various namespaces will not be conflated. For example, a "table" element could be defined in a new namespace with new semantics different from the HTML "table" element and the browser will be able to differentiate between the two. In providing namespaces, XHTML combined with CSS allow authoring communities to easily extend the semantic vocabulary of documents. This accommodates the use of proprietary elements so long as those elements can be presented to the intended audience through complete style sheet definitions (including aural/speech and tactile styles).

Note that XHTML is parsed as XML by major web browsers only if it is served using the MIME type. Many XHTML documents are currently served on the Web using the MIME type, in order to ensure backwards compatibility with older browsers, and with all current releases of Microsoft Internet Explorer, thereby making losing the possible gain of using XHTML. See also the discussion of this issue in the XHTML article.

HTML5
HTML5, currently in draft status only, aims to be the most complete solution to the problem of tag soup thus far while remaining as backwards- and forwards-compatible as possible. By contrast to XHTML, which departs from backwards compatibility and takes the approach that parsers should become less tolerant of badly-formed markup, HTML5 acknowledges that badly-formed HTML code already exists in large quantities and will probably continue to be used, and takes the view that the specification should be expanded to ensure maximum compatibility with such code.

Thus, the HTML 5 specification has altered its definition of HTML syntax both to accommodate common syntax in use today, and to explicitly describe exactly how "badly-formed code" should be treated by the parser. The handling of badly-formed code now has a place in the specification itself, hopefully reducing the need for future HTML parsers to implement additional, out-of-specification measures for dealing with code that it does not recognize.

Tools to fix tag soup

 * HTML Tidy is a software tool available for many platforms which can correct invalid syntax, and most invalid document structure, outputting HTML-like code to HTML or XHTML.
 * Aggiorno is a Visual Studio add-in that focuses on making web sites standards compliant
 * Tagsoup is a Java library that parses HTML, cleans it up, and delivers a stream of SAX events representing well-formed and valid XHTML