Printing XML
Why CSS is better than XSL

Who am I

• Michael Day, mikeday@yeslogic.com
• CEO, YesLogic Pty Ltd
• Melbourne, Australia
• architect of Prince XML + CSS formatter

What is Prince

• XML + CSS → PDF
• XML: XHTML, DocBook, or whatever
• CSS: CSS 2.1 as used on web, also some CSS 3
• we support SVG and are working on MathML

What can Prince do

• this book!
• written in XHTML
• styled with CSS
• converted to PDF with Prince
• everything a book needs
  • page headers/footers duplex layout, page numbers
  • floats, tables, images
  • table of contents with cross-references and leaders

Why am I here

• Why did we make Prince?
• Two beliefs:
  • XML is a great way to represent documents
  • CSS is a great way for styling XML for printing
• Prince is the result of these beliefs: printing XML with CSS
• not going to talk about first belief, take it as given
• second belief needs more explanation

Is CSS better than XSL?

• provocative topic, sounds somewhat confrontational
• overlap between two approaches, both attack similar problems
• CSS is more economical
  • less time, less money, less stress
• if you can do it with CSS, you can probably do it with XSL
  • more time, more expertise, more maintenance
• people tell us the CSS approach makes their life easier

Quick CSS intro

• rules, selectors, properties, values
• concise syntax, human readable and writable

CSS is easy to learn

• h1 { color: red }
• very low barrier to entry
• success on the web

Quick XSL intro

• XSL = XSLT + XSL-FO
• XSLT: XML scripting language for transforming XML documents
• XSL-FO: presentational markup
• XSL approach uses XSLT to transform documents into XSL-FO

XSL is a challenge to learn

• requires mastering two new XML vocabularies
• requires in-depth knowledge of XML data models
• "style sheets" are compound docs with three namespaces
  • your input document namespace
  • XSLT elements
  • XSL-FO elements
• training issues
  • so much to learn
  • who will learn it? programmers? designers?

CSS is based on style rules

• property-level granularity
  • allows arbitrary division / composition of style sheets
  • promotes modularity
  • saves typing by reducing duplication
• example:
  • rules for fonts, colors, margins
  • rules for headings, paragraphs, quotes
  • flexibility in style sheet structure

CSS style rules are declarative

• declare the style you want, CSS makes it happen
• * { text-underline: none }
• *.legalese { text-transform: uppercase }
• h1 + p { font-size: 125% }
• a[href]::after { content: attr(href) }
• CSS rules are declarative
  • they declare the desired style
  • they make perfect sense in isolation
  • CSS implementation is responsible for cascading process

XSL is based on element templates

• exactly one template matches each input element
• templates emit output elements/attributes
• default template copies text recursively
• element-level granularity

XSL templates are not declarative

• aha you say, XSL is declarative too!
• yes, but... everything is declarative
• question is, what level of abstraction?
• even assembly language is declarative
• XSL is declarative language for XML transformation
• XSL is not a declarative language for styling

Why is this a problem

• earlier examples cannot be done in XSL
  • * { text-underline: none }
    • requires modifying every template
    • alternatively, implement second transform pass
  • *.legalese { text-transform: uppercase }
    • requires duplicating every template
    • or adding if test or function call to every template
    • alternatively, implement second transform pass
  • h1 + p { font-size: 125% }
  • a[href]::after { content: attr(href) }
    • existing templates must be modified

What is XSLT

• XSLT is a programming language
  • parameters, variables
  • conditional logic
  • recursive functions
  • external scripts
  • Turing complete
• XSLT is an XML scripting language
  • equivalent of grep/awk/sed/perl for XML world
  • handy for writing transforms and little XML utilities
• XSLT 2.0 reinforces this by adding features like typing

XSLT is unsuitable for styling

• XSLT is not a styling language!
• domain of XSLT is elements, attributes, templates and functions
• perl scripts that transform XML are not style sheets
• writing programs to format documents is suboptimal
  • requiring programming impedes designers
  • higher maintenance cost
  • no separation of concerns: content manipulation and styling

XSLT is still useful

• publishing pipelines often need more than just styling
• small modular transforms
  • generate table of contents, index, glossary
  • sum invoice totals
  • perform macro pre-processing: <logo/> → SVG
• many other processing steps:
  • XInclude processing
  • schema validation / augmentation
  • database queries to fill in values
• XSLT has role to play as one tool among many

CSS is capable of printing

• page style
• page breaking
• page headers/footers
• page numbering
• named pages
• duplex printing
• cross-references
• leaders

CSS enables content repurposing

• CSS works across many media out of the box
  • browsers on screens, presentations, printing, aural
• @media rules
• combine specific style sheets
  • common.css
  • screen.css
  • print.css

Why not use CSS for printing

• CSS has had a huge impact on the web
  • more styling possibilities
  • cleaner markup
  • greater range of dynamic applications
• focus on webpages means that printing is often neglected
  • browsers concentrating on screen media
  • continuous, not paged layout
  • printing lags behind, even though CSS supports it
• duplicating web content for printing is a hassle
• CSS is good for printing, everyone knows it, why not use it?

CSS printing tools

• most tools that print don't do much CSS and vice versa
  • browsers, html2ps, LaTeX converters
• Prince has first-class CSS support with emphasis on printing
• Prince 5.0 beta release last week
  • PDF links/bookmarks/security
  • support for wide range of UNICODE
  • CSS positioning
  • MacOS X support
  • many other new features and improvements

The more the merrier

• we want to be the best out of a strong field
• we encourage XSL-FO formatters to support CSS
  • then can truly say you support printing XML
  • many of your users are frustrated, we hear them
  • promote standard models for styling, printing