Against Markdown
14 points by abhin4v
14 points by abhin4v
Markdown won because it's good enough and everyone already knows it. Perfect is the enemy of shipped. I'm not rewriting my docs in HTML to gain semantic purity that 99% of users will never notice.
Markdown won because it's good enough and everyone already knows it. Perfect is the enemy of shipped.
You don't really need to invoke "worse is better" to defend Markdown.
"Worse is better" is a much more controversial principle than the simple fact that there isn't a "better" in the first place — there is nothing to be gained by replacing Markdown, as you correctly noted. It is handy, easy to use, and covers >90% of technical writing needs. There is simply no point in making one's life harder for abstract semantic purity.
Markdown is easier to learn for someone who is not very technical, making it a great alternative to WYSIWYG editors. While it is far from perfect, it is the tool that covers 90% of needs without requiring too much effort.
I've seen this point raised many times, but is this actually true? Intuitively it makes sense to me, but anecdotally I don't feel like that is the case. I can say that when I observed less tech-savvy people editing a markdown-based wiki they stuck mostly to unformatted text, and even in slack that has a WYSIWYG markdown editor I feel like most people just forego using it. At least in the first case text formatting is something that would have been used if it were easily accessible to them.
There is no way I'm going to write my docs in HTML and I'm a person who is able to do that and has done so in the past. It was absolutely terrible.
If anything, we should be more for markdown.
What's the use of going through the effort of coding semantic content in the age of LLMs? Surely an LLM can accurately identify Gone With the Wind as a title, raison d'être as a loanword, and fabulous as an emphasized word.
are you suggesting that instead of markdown we write without markup and let an LLM convert it to HTML?
No, I'm saying that there's no need to include semantic differences like (hypothetical) <title>, <loanword> or <cite> in HTML, so the fact that Markdown just converts _italics_ to <em> and __bold__ to <strong> is not a big problem.
I read markdown in plain text all the time, and I specifically write it to be effectively read via source (numbering numbered lists instead of allowing auto numbering in renderers, properly spacing tables and putting effort into keeping them narrow).
The only really ugly thing about reading my markdown source is I'm a big fan of semantic line breaks, so reading paragraphs in source code is a little weird.
I like reStructuredText and since the Linux kernel chose it; I'd say it won. Has an escape sequence, and the reference parser is python's docutil.
I took a look at that page and the Wikipedia article in ReST, and I notice the testimonials page hasn't been updated since 2002, and there's not many implementations of ReST other than the reference implementation and Pandoc.
Part of Markdown's appeal is the wealth of implementations. I managed to speed up my blog generation by a very large amount by implementing CommonMark's C library.
I also don't use Setex type headers, I prefer ATX style.
Over the years I wrote documents in POD, LaTeX, Typst, HTML, t2t, roff/mdoc, djot and markdown. Markdown (or it's modern reincarnation djot) is probably the only one I'm able to use without a search engine. That's the main reason I'm using it despite all it's flows.