By John Gruber
Tara AI — Build better software, faster
“Sometimes the truth of a thing is not so much in the think of it, but in the feel of it.” —Stanley Kubrick
Here’s a question: When was the last time you listened to an argument, and on the basis of that argument, changed your mind? Not just about something you hadn’t really given much thought to, but something which, prior to considering the argument in question, you felt quite certain regarding your original stance.
In other words, when was the last time you realized you were completely wrong on a matter of opinion?
If your answer is “never”, or even “a long time ago”, is it because you’re always right?
Here’s one of mine.
Back in January, there was a debate regarding the application of Postel’s Law in software that parses XML syndicated feeds (formats such as RSS and Atom). The nutshell synopsis is that Postel’s Law states: “be conservative in what you do, be liberal in what you accept from others”; but the XML specification clearly states: “Once a fatal error is detected, however, the processor MUST NOT continue normal processing (i.e., it MUST NOT continue to pass character data and information about the document’s logical structure to the application in the normal way).”
So, what should software that parses XML syndicated feeds do? Follow Postel’s Law, and be liberal about errors in a feed? Or follow the XML spec, and stop processing the feed at the first fatal error?
Brent Simmons (developer of NetNewsWire, the leading Mac OS syndicated newsreader) and Nick Bradbury (developer of FeedDemon, the leading Windows syndicated newsreader) both decided that their software would be strict when parsing Atom XML feeds. Many other smart people agreed with them.
I think it’s fair to say the “clients should be strict” argument boiled down to this:
Valid, well-formed XML is better than invalid XML.
It’s not that hard to write software that produces valid, well-formed XML. (Or in the words of Tim Bray: “Anyone who can’t make a syndication feed that’s well-formed XML is an incompetent fool.”)
If the leading client apps for consuming XML feeds required valid and well-formed XML, it would put pressure on apps that produce feeds not to generate bad XML.
I agree with all three points, and thus, I was firmly in the “clients ought to be strict about parsing” camp.
But then I read Mark Pilgrim’s “Thought Experiment”. Pilgrim is not just a proponent of liberal feed parsing — he’s put his code where his mouth is and written the very-well-regarded open source Universal Feed Parser.
Pilgrim’s key point, at least as I saw it, was this: if you’re writing software that consumes XML feeds, and your parser isn’t at least somewhat liberal, your users will suffer when they encounter a malformed feed. And, eventually, your users will encounter malformed feeds. When that happens, the producer of the broken feed may well be at fault, but it’s the users of your software who will suffer by enforced strictness at the client end.
Solely on the basis of his argument in “Thought Experiment”, Pilgrim persuaded me that I was wrong. But what’s interesting is that in doing so, he did not refute a single one of the three points that led me to side with the “be strict” camp in the first place.
Thus, Point No. 1: When I end up changing my mind on a matter of opinion, it usually is not because I had the facts wrong; it’s because I was looking at the wrong facts.
The basic idea behind all weblog software — reduced to a nutshell — seems so simple in retrospect. Rather than managing a web site as a collection of pages, you manage a collection of posts, and the weblog software takes care of turning the posts into pages for you.
The appeal of weblog software for non-nerds is obvious. Without it, they wouldn’t be able to publish their sites.
But what’s the appeal of weblog software to nerds — HTML experts who are fully capable of hand-coding a weblog with HTML files? Sure, there are a few hand-coding holdouts, but for the most part, even the most knowledgeable and renowned web developers in the world are using weblogging software packages. (Or they’ve written their own publishing software.)
The answer is convenience and flexibility. Weblog software takes away an inordinate amount of the monotony involved with updating a web site. I must profess — I personally didn’t figure this out until 2002, a few months before I launched Daring Fireball. The fact that I was capable of hand-coding an entire web site — and in fact considered hand-coding easy — blinded me to the fact that it involved an awful lot of repetitive monkey work.
Here’s what happens every time I post a new article to Daring Fireball:
I take just one action — posting a new article — and Movable Type creates one new file and updates four others. None of these tasks would be difficult to accomplish by hand. Mostly, it’d be a copy-and-paste job. But it sure would be monotonous — and even though it admittedly would not be hard, there would still be a high chance I’d make mistakes. I’d make them because I’d get bored or simply forget to perform one of the steps. Monotonous repetitive tasks are exactly what computers are good at, and what humans are bad at.
Plus, the encapsulation feels natural. Writing a new article — or making changes to an existing one — feels like it ought to be one task. It’s the article I want to edit, not the individual pages where it appears.
Thus, Point No. 2: Just because something isn’t difficult doesn’t mean it’s the way it ought to be done.
Let’s take a step back. Above, I wrote that weblog software allows you to manage a web site as a collection of posts instead of pages.
But so what is a post?
The conventional wisdom is that a post is a snippet of HTML. Not a
full HTML document, just a snippet of HTML-formatted text — the
weblog software takes care of forming full HTML (and/or XML)
documents. Your weblog templates contain all the tags related to
document structure —
<body>, etc. — and they
have slots for where your posts will go. When you publish, your weblog
software takes your posts, as snippets of HTML, and plugs them into
your HTML templates.
For the full first year I published Daring Fireball, from August 2002 to around August of last year, I was perfectly happy to buy into this post-as-snippet-of-HTML concept. I never really even considered an alternative. Everything I wrote for Daring Fireball, I wrote formatted as valid HTML. (Actually XHTML, but the difference is irrelevant here.)
Except, of course, snippets of HTML can’t be validated, because HTML is a document format. You can write well-formed snippets of HTML — making sure you close all your tags and escape all your ampersands and angle brackets — but you can’t pass a snippet of HTML to the W3C HTML Validator, nor can you pass it through BBEdit’s HTML syntax checker.
My intended workflow:
But my actual workflow looked like this:
Eventually, it dawned on me: this is madness. The primary advantage to using a computer for writing is the immediacy of editing. Write, read, revise, all in the same window, all in the same mode.
The argument for writing text in raw HTML — the argument I used myself for years — is that HTML is not hard. I still agree with this. HTML is pretty easy to learn, and once learned, easy to apply. Full-blown web development? That’s hard. But the basic tags and rules of HTML — enough HTML to be capable of composing weblog entries in raw HTML — that’s easy.
But there’s a reason why plain text browsers like Lynx don’t just show you the raw HTML source code. It’s simply not meant to be a readable format. Doesn’t it strike you as odd to write in a format that isn’t readable? Suddenly, it struck me as absurd.
Application of Point No. 1: I’m not arguing that raw HTML is hard. I agree that it’s easy. I’m arguing an orthogonal point — that what it’s easy for is tagging a document with markup, not reading and composing prose.
Application of Point No. 2: Even if you insist that it’s easy to
compose text in raw HTML, isn’t it a chore? Isn’t it tedious to write
AT&T” instead of just “
AT&T”? (Not to mention the necessity
of encoding ampersands within URLs?)
It’s 2004. Shouldn’t your computer be able to determine where you’ve written paragraphs and sub-heads?
And don’t tell me that Movable Type’s “Convert Line Breaks” feature does that for you. MT 2.661’s “Convert Line Breaks” will turn these two lines of input:
<h2>This is a header.</h2> This is a paragraph.
<p><h2>This is a header.</h2></p> <p>This is a paragraph.</p>
Which is just silly. TypePad is apparently smart enough not to wrap
<p> tags around block-level HTML tags — so MT 3.0
probably will be too — but it still doesn’t alleviate any of the
other tedium or visual clutter of raw HTML.
Why is it that desktop weblog editors need to provide “preview” modes? You don’t need to “preview” an email before you send it — you write it, you read it, you edit it, right there.
In fact, I love writing email. Email is my favorite writing medium. I’ve sent over 16,000 emails in the last five years. The conventions of plain text email allow me to express myself clearly and precisely, without ever getting in my way.
Thus, Markdown. Email-style writing for the web.
Most other text-to-HTML filters are based on the premise that HTML tags are difficult, and so they go out of their way to replace HTML tags with their own, which end up being neither “easier” nor more readable than HTML. And at the same time, they end up making it hard to drop into manual and just use raw HTML when you really need to.
Other filters are aimed at replacing HTML. Markdown is aimed elsewhere. It’s aimed at a sweet spot, between making it easy to use real HTML when you need it, and letting you just write plain text for anything where it’s sensible and obvious to do so.
The tags that most weblog apps provide shortcuts for — italics, bold, links, paragraphs, blockquotes — are the ones you shouldn’t need to worry about “tagging” in the first place. Making it easy to insert these tags does nothing to make it easier to write, and worse, makes your composition harder to read.
But when you do need to use inline raw HTML — say, to create a specially-formatted ordered list using custom class attributes — you ought to be able to just drop into HTML. No escapes, no special mode-switching markers, just use the tags. Markdown lets you do this, because it’s designed specifically and only as a pre-processor for HTML.
(If you really do want to translate a Markdown-formatted document to some non-HTML format, just translate it to HTML first, then use an existing HTML-to-whatever filter.)
And while I do think HTML is easy, there’s one particular area where
it is in fact quite tricky, if not downright difficult: using HTML
markup to write about HTML markup is a major pain in the ass. When
you write about code, you should only have to worry about the example
code itself — not about escaping every single instance of
In addition to making it easy to include inline HTML tags in a document, Markdown also makes it easy to include example HTML tags within code spans and code blocks.
The typographic constraints of plain text — a single typeface, in a single size, with no true italics or bold — are very much similar to the constraints of a typewriter. Imagine that someone was nice enough to buy you a gift: an original typewritten manuscript for a classic novel. Let’s say Fitzgerald’s “The Great Gatsby”. You could sit down with this manuscript and read it, straight through, and get pretty much the same reading experience as you would when reading it in the form of a nicely bound and typeset book. Yes, it would all be set in the typewriter’s smudgy fixed-width Courier-esque typeface, with underlining instead of italics, etc. — but the words would still flow, from page to mind, just as Fitzgerald intended.
The quote from Stanley Kubrick I used to start this article is one of my very favorites. When you write and read text that’s marked-up with HTML tags, it’s forcing you to concentrate on the think of it. It’s the feel of it that I want Markdown-formatted text to convey.