Slashdot Mirror
GNU TeXmacs and Structured Text Editing
Joris van der Hoeven writes "It is a common belief that structured texts are best conceived using ASCII-based text editors like Emacs or VI. It is true that word processors like MS-Word have done a bad job on this issue. But does this mean that wysiwyg structured text editing would be impossible? We firmly believe the contrary and
argue
that such editors are both technically conceivable and desirable. Judge for yourself by taking a look at the GNU TeXmacs
program, whose version 1.0 has just been released."
Good question. LyX is great. I wrote a book typeset directly from LyX-LaTeX-PDF.
Judging from the screen shots on the web site, TeXmacs is more oriented to mathematical writing while LyX is more for general typesetting. The TeXmacs icons on the tool bar also suggests this.
It suffers from the same ailment (although far less so) as HTML; The layout is intimately linked with the content. As long as font size information, background color, text alignment, etc. are part of the document and mixed with section, paragraph, bibliography, etc. there will be trouble.
This is precisely why things like DocBook came into being. It contains absolutely no layout information. It is all about structured content. Layout is handled later by a separate processor. This does not necessarily mean that input must be so sterile. In fact, I believe that a WYSIWYG DocBook editor would be a godsend in providing a "way out" for all of those Word authors but still applying content structure. Perhaps a variation on the Mozilla Composer concept?
And for those who will let go of their emacs when you pry it from their cold, dead fingers, there is at least one XML editor that takes DTDs for input to aid in tag creation and allows for hooks into XSLT processors for "pretty" previews. It called XAE (XML Authoring Environment for Emacs).
- I don't need to go outside, my CRT tan'll do me just fine.
But TeX enthusiasts seem to be stuck on the idea that Tex is also useful for structured documents. Sorry, it just isn't. If you want to impose structure on a document, you can't use a format designed around layout. Even if you add constructs that describe document structure (as LaTeX does) you can't prevent the user from using non-structure elements "because it looks right". So you end up with a convoluted mixture of structure and layout that's impossible to maintain. That's why HTML is such a mess. That's why maintaining large technical documents with traditional word processors is a nightmare.
If you need to maintain a large structured document, you need to use a format that makes no attempt at all to describe layout. So the writer is forced to think purely in terms of how the document is organized. You keep layout description in a separate thing, a "style sheet". Not only does that end your document maintainence nightmare, but it allows you to deliver the same document in different ways just by providing the appropriate style sheet. You have a single source that's accessible as a set of web page, or as a printed document, or whatever.
What formats am I talking about? Since this is 2002, I'm talking about XML. Not XML in general (most XML apps are data-centric not doc-centric) but specific appropriate XML applications, such as DocBook or DITA. For the stylesheets there's Cascading Style Sheets and/or XSL. But these are just the best technologies that happen to available now. The basic idea has been around for a long time: in structured documents you have to separate markup and layout.
I think this GNU thing suffers from the same problems as LyX, Scientific Workplace, and every other GUI front-end to (La)TeX -- it relies on menus and the mouse.
In the time it takes someone to remove his hand from the keyboard, search through a menu or click a button to make a fraction, and search through another menu or palette to find a gamma, I could have typed \frac{\gamma}{2} ten times.
My experience has been that people look to these things so they don't have to be bothered by knowing all the commands. I think that's a waste of time. After writing one paper using LaTeX, you will have memorized all of the symbols commonly used in your discipline, and you'll soon discover that LaTeX is so much faster than a GUI application.
Spending a couple of hours learning LaTeX is time well spent, and you will certainly be repaid many times over in the long run.
Sig (appended to the end of comments you post, 120 chars)
But, in the end, it is the layout that matters, is it not?
.net</a></div>
Don't worry, I was like you too once. I beleived that it was possible to define a document format so that I could separate the "look" of the document with the actual information that I was trying to convey. But it turns out that this only works with simple documents. As your document increases in complexity, you shouldn't need to define new markup to make the document truly structured and portable. So what has to happen, and it does happen all the time, is that the document author goes beyond the markup and considers the presentation of the document.
You don't believe me? Take a simple example, I'm rewriting my homepage. On the first page, I want to put my email address and I want it to be set apart from the rest of the page. This is what I have in html:
<div><b>Email:</b> <a href="mailto:kholmes@sedona.net">kholmes@sedona
Div is a generic tag used for block elements. Why did I use it? Its definitely against the structured document approach. Otherwise I could simply use div tags for the entire document. Well, the above is certainly not a paragraph. Even if I was defining my own markup language for XML, there is no word to describe it. I could create a specific tag <email-heading>, but again this goes against the structured document approach. Specific tags are not generally useful.
The problem is that for structured documents to work, you need to *write* in a structured fashion. You have to keep track of what the semantics of each part of the document is called and be sure to limit yourself to these semantics.
Any writer would know that this would be far too limiting. Writing, inevitably, is a right-brain activity and is fundamentally unstructured. Technical documentation has to more structured than most since it needs to be easily referenced but inevitably the technical writer is plagued with the same problems that what is natural to write is not structured well to markup.
Then, when you leave the arena of technical documentation structured writing becomes almost pointless. Newspaper columns and magazine articles are really just a sequence of paragraphs written after each other. There is really no need for markup at all. Markup is barely useful for writing fiction since it comes in so many forms. And if the author chooses a new form you can't say "Wait now, we need to write markup for that." When in the end, the author is writing to a reader and not to a computer. Presentation is important. And if you intend on using a structured editor for typesetting poetry, I think even the most stringent holdouts would agree that this is a hopeless cause.
When you define a markup with a DTD or XML schema, you are saying that these are the only things you may write. To write new semantics, the writer must in his head determine the presentation.
Presentation can not be separate from content.