WWYW2SWYGWINDWY — the case against WYSIWYG

October 19th, 2006

Try to imagine Jane Austen, Friedrich Nietzsche, or Victor Hugo starting a new work: perhaps they would choose a special handwriting style, they would order a special paper format, they would bind it with the best leather and thread. Then they’d start writing, meticulously arranging letters so that the last word of each line would exactly touch the right margin, patiently underlining headings with a wooden ruler and a steel pen dipped in purple china ink, sharpening their pencils for a finest point in order to be able to write the wee letters in the footnotes. In other words, making all sorts of childish and pointless efforts so that their manuscript mildly resembles a printed book.

Or maybe they’d just write

I can hear your smile imagining poor Miss Austen painfully tracing navy-blue block capitals with a fancy pen for chapter headings. You know she would not do that. A manuscript is a manuscript. One writes the content. Only when it’s finished writing, one works with an editor to markup the manuscript: this is a chapter heading, this is a quote, this is a footnote, etc. Then one agrees with a typesetter what fonts will look best, what paper size will improve legibility, should quotes be italicized, etc.

If everybody knows this, then why is it that most people do this backwards and use WYSIWYG ‘word processing’ software, and especially that one which-must-not-be-named ? Why do you bother with boldface headings, justified text, headers and footers, manual pagebreaks, page margins and all that sort of things when you have not finished (or even not started) writing ?

WWYW2SWYGWINDWY? As if anyone needs to see the finished work building itself as one is trudging along the creation process… Did Shakespeare need to have real actors read out his lines as he was crafting them ? Pen and paper was enough. Did Mozart need to have an orchestra playing his notes as they came to him ? All he used was a piano. Did William Lamb need to have workers assembling steel trusses before his eyes as he was designing the Empire State Building ? He could do with blueprints. I am no Shakespeare, Mozart, Lamb, but I can tell you a text editor (with a spell checker) is all I will ever need for writing.

Once the writing is done: markup languages

Markup languages are the digital equivalent of red-ink analog markup: once your manuscript is finished (or before, if you absolutely need to have a glimpse at a finished product), you use formal markup to define the layout. SGML, HTML, XML or LaTeX, they all have the same purpose: you do not specify how things will look like, you tell the computer what is what.

Basically, instead of writing:

Of this at least, he was sure (<-- this last word should be emphasized according to any general convention and particular style that the typesetter will see fit, although I'd prefer boldface myself, but what do I know about it)

You very conveniently write:

Of this at least, he was \emph{sure} (LaTeX syntax)

Then somewhere else, there is a document specifying how markup indications should be interpreted. It can be called a stylesheet, a schema, a documentclass, but it serves the same purpose of containing the set of rules that a typesetting engine should follow to apply layout to the text.
As a matter of fact, as markup languages are based on plain text, they can be read by human eyes: you could give an XML manuscript to a XIXth century typesetter, and if he had the corresponding stylesheet to determine layout rules, he could make a fine printed book out of it.

Markup and legibility

As there are more and more indications in a marked-up manuscript (paper or digital), it get harder to read for further editing. I can think of five options to work with such documents (in increasing order of technology burden):

  1. Get used to the markup and continue working with the cluttered manuscript — this is my choice.
  2. Use a small utility that strips the manuscript from all its markup. Unfortunately, you lose the layout indications.
  3. Use an interpreter to display the final result in a separate window (e.g. in a web browser), while you continue to work in an editor with the plain document: with a little training, you’ll find your way around the manuscript quite easily. For instance, working with XML, just hit ctrl-s, alt-tab, ctrl-r when you want to see your changes.
  4. Use a WYSIWYM editor (I know at least one: LyX, to edit LaTeX documents). The principle of those is to display the text without the markup, but with a simplified layout, just enough that you know without a doubt what markup is behind.
  5. Do NOT use a WYSIWYG editor.

The curse of WYSIWYG editors

Let me explain the last recommendation: the fundamental curse with any WYSIWYG editor is that you can never be sure what markup instructions are used to format the text. For instance, to tell apart heading 4 from just Normal + flush left + 14 pt + bold + space above: 6 pt, you have to select the line and check the style definition. In the former case, the software knows this is a heading of level 4: when you change the layout for such headings, or you choose to display them in the table of content, or you promote the whole section by one level, the changes will be propagated. In the latter case, although it looks just like a heading 4 (thus tricking you into believing that the computer knows what it is), it will be left behind miserably. And then I can hear the sound of hair being pulled.

In addition, seeing what you get often distracts you from the true task of writing: have you ever strayed to change a word because the hyphenation made the justified paragraph look funky ? to refactor a sentence because of an odd pagebreak ? to change the ‘heading 2′ style because a chapter title did not fit on one line ?

And then, there are the technological issues. As per design, WYSIWYG editors need to process the layout in real-time. Therefore, past a certain manuscript size (typically beyond 100 pages with illustrations), they begin to sink under the load. There are generally ways to split the document into sub-documents for a more efficient processing, but I know few people using the functionality, and in any case you will have to merge sub-documents in the end to resolve cross-references and page numbers. Chances are that your computer will crash under the burden just when you are in a real hurry to get your document out in a printed form (paper or pdf).

On top of all this, if you combine the curse of WYSIWYG with the shortcomings of that one editor which-should-not-be-named (e.g. file corruption, backwards compatibility of proprietary binary formats, cross-platform issues, bad standards, cost), you end up with a tool that is just as appropriate for writing as a scented sledgehammer or a dead pink cow.

As a conclusion

In my opinion, a marked-up plain-text manuscript is the source code, the program, the archdefinition of what the author intends. A WYSIWYG document is generally no better than a xerox copy of the text: you see the result, but the intention is lost.

Epilogue

If you really must continue using WYSIWYG word-processors, please stick to a strict stylesheet: banish the ‘boldface’,'italics’, or ‘bullet list’ and ‘indent’ buttons from the toolbar. And always prune the stylesheet to make sure it does not contain more than 10-20 styles.

Read on

What I know best
P-books

Read away

Largely inspired by this article
LaTeX on Wikipedia
WYSIWYG on wikipedia
Markup languages

5 Responses to “WWYW2SWYGWINDWY — the case against WYSIWYG”

  1. LK Says:

    And with paper being expensive probably and with no white-out available (or delete button), they had to be pretty careful with their words

  2. mandarine Says:

    I do not know about the XVIIIth or XIXth century, but back in medieval times, paper or parchment was so expensive that monks bleached ‘lesser’ books to write new things on them. These palimpsests can now be deciphered with modern techniques, so that long lost works once considered ‘lesser’ can be retrieved from oblivion.

    I wonder whether future techniques will be able to undig all the layers of overwritten data from our hard disks.

  3. DevNull Says:

    Nice article. Now I have to find a dead pink cow so that I can convince Word users that it’s not the best tool :)

  4. Rustum Says:

    Great blog. I found it via the p-books article, following your comment at Steven Poole’s blog.

    I sometimes mess around with minimalist word-processors, but have not yet decided whether or which to use. As far as WYSIWIG word-processors go, I have one font I use, I don’t use stylesheets apart from the normal, and keep everything at its simplest level, typing until I am done before formatting.

    In my academic days I used WordPerfect 6 and produced a 220 page document with it, with notes and bibliography, but avoided sub-documents and merging. WordPerfect was brilliant because of the way it implemented style codes, and because one could easily sort out problems through reveal codes.

    In any case, I am enjoying your blog, but am a little bothered by the fact that there are no dates for posts. When I like a blog, I spend quite a lot of time reading, and like the historical frame that dates provide.

  5. mandarine Says:

    You are absolutely right about the date thing. I had removed them some time ago thinking that it would not be a problem, as the posts I write tend to be timeless. And now I find I do care myself and am the first one to be bothered not to know when I wrote this or that.

    So you win: I have just put the posting dates back.

    Let me know if you really must have the commenting dates too.

Leave a Reply