I have written a short note about using
sed with Markdown
to maintain a collection of HTML documents. Programmers are forever writing
documentation (proposals, specifications, technical notes, and so on),
and the latest stage in my quest to make writing prose as frictionless
I have started using Markdown to do this.
Before that I wrote in HTML with some XML mixed in to better handle
sectioning; before that I used Microsoft Word.
The aformentioned note gives the details of how this works. The rest of this note discusses why I'm not using Microsoft Word.
Writing with Word is not frictionless
Most of my managers would expect me to use Microsoft Word to write proposals and specifications and the like. This makes some sense because, not being developers, the spend most of their time plugging away in Microsoft Office, so Word is always to hand. But I instead program for a living: I typically have three text editors open at once (Microsoft Visual Studio .NET 2003, the Pythonwin IDE, and jEdit), so dashing out a note in a text editor is the natural thing to do. When I do use Word I waste time struggling against its officious auto-lousification features. Jeremy's description of it as being like a bolshy secretary is very apt; it is so annoying that there is a whole industry growing around the Word Annoyances books.
Besides, we are all web developers now, and should always be looking for a chance to sharpen our HTML skills and practice organizing documents for the Web.
Word documents are not good to read
Word is worse for reading on-line than HTML. When colleagues and customers send us documents in Word format, I cringe. I have to wait while the word processor drags its bloated code off the hard drive and in to memory, and eventually opens a window in which the document in question is displayed at some random zoom level and view mode chosen by the last person to edit it.
Word documents do not sit comfortably on the company intranet. Not that anyone seems to have the wit to appreciate that being able to refer to documents by URL would save a lot of time that instead we waste digging though old mail messages or obscure file shares looking for 'the DTI spec' (which is not named that, of course).
With HTML pages on an intranet server, it is easy to have cross-references between documents, to write index pages, even share them with clients via an extranet server. You can have links from bug reports to the relevant section of the specification or what-have-you.
Word documents are ugly
But surely we can exploit the extra formatting features Word offers to make documents of rare clarity and great beauty? Yes, and, back when I wrote position papers in Word, I carefully chose fonts and adjusted the layout of headings to make the documents look nice and read well. I even persuaded the powers at be to let me buy a Bitstream 500-font CD so we could use Plantin rather than Times Roman. I made these changes using Styles so that formatting was consistent throughout the document. I discovered how to make a Code style that used Univers 55 to represent computer text and suppressed the spell-checker. I even worked out how to attach this style to a keyboard shortcut so I could write technical documents with lots of computer text in. I spent ages arguing with the automatic header-numbering dialogues until they numbered the headings properly, and I tweaked the table of contents to make it more usable. I organized my documents carefully so as to avoid convoluted and heavily nested sub-sub-subheadings. I felt this was the least I needed to do to produce documents of an acceptable standard that could be shown to clients.
In this regard, I am unique. Normal people use the wizz-bang features of Word to create ugly, unstructured documents that use fonts too small to read on the screen, and lines too long to read comfortably on paper. The least bad examples are when they leave the default, ugly, heading formats in place. When they attempt impose a style of their own it is far worse.
Concentrate on the words, not Word
Another things that irritates me (and is not entirely relevant to this discussion) is that so little thought goes in to writing these words. Another way in which I am freakishly divergent from the norm is that the second thing I put in every paper is a line with the revision number, the date, and my name. The idea is that (a) it should be easy to discover whether one is reading the latest version, and (b) if there are questions, it is possible to ask someone. An undated document headed 'Bugs in Frobomatic' followed by a list of complaints will soon be useless because it does not say which version of the software, or whom to ask for details about the bugs.
I used to make a point of numbering documents (something might be Tech Note #3) and number headings in documents and in some cases even paragraphs. This way you could refer to TN3, §4.2 concisely rather than 'The one about the database thingummy that Jared mailed you in June I think, about halfway through, headed Issues with User Authentication (or do I mean Authorization?), about four paragraphs in'. I even wrote scripts to automagically number headings in HTML documents, and provide anchors so I could link directly to a certain section of a certain document. I have since given up: nobody cares.
Too many people spend hundreds of pounds on word-processing software so that they may spend hundreds of hours writing documents, but then neglect to spend a minute or two ensuring these documents will be citeable, findable and readable.
For writing technical prose, the formatting features of word processors are a distraction at best. Programmers especially would be better off spending their time getting really good with a decent text editor.