The fact that this is an interesting article is a huge indictment of CSS. It was developed for a different, simpler use case, and it's obviously the wrong abstraction at this point. Too bad it's entrenched.
Pragmatically: I suppose that we embrace CSS/HTML/JS as object code. And I mean this in a much more dramatic way than CoffeeScript or SASS. This is finally starting to happen, thanks to browser support for things like source-maps. This is one reason I contribute to ClojureScript!
Idealistically: We should be pushing for lower level platforms for code distribution and execution, as well as host platform integration. Portable Native Client and similar efforts shouldn't be so controversial.
There have been improvements in CSS2 and CSS3, like adding table layout so people could at least use the table layout features on div elements rather than abusing tables for layout, and CSS flexbox that work well for standard UI widget layouts.
But the big fixes to layout have been harder to specify and agree upon. They've been in the works for many years, but it's taken a while to figure out how they really should work, and which parts should be fundamental. It looks like it's getting a lot closer; the CSS Grid Module http://dev.w3.org/csswg/css-grid/ is being treated as the fundamental part of the stack on which others will build, it is being worked on by several browser vendors and is in highly active development, and several browsers have implementations in development. It seems to draw elements from IE's experimental CSS grid, as well as the templates of the earlier template layout proposal.
So, yeah, for many years CSS has had a big deficiency in specifying page layout. It's going to get better very soon, however, once the CSS Grid Module is finished and implemented. I'd imagine we'll start seeing implementations within a year; here's Webkit's tracking bug for it: https://bugs.webkit.org/showdependencytree.cgi?id=60731&hide... here's Chromium's: https://code.google.com/p/chromium/issues/detail?id=79180 . Mozilla doesn't seem to be working on it yet, though they have had some preliminary discussion: https://bugzilla.mozilla.org/show_bug.cgi?id=616605 . IE had implemented an experimental Grid Layout proposal that the current work is based on, so while the syntax won't be exactly the same, they have most of the underlying functionality working: http://msdn.microsoft.com/en-us/library/ie/hh673533(v=vs.85)...
Although lower-level and not as modular, flexbox can already be used with great success for a number of layout sections (media objects, vertical centering, input inserts) or full layouts ("holy grail", sticky footers and grids): http://philipwalton.github.io/solved-by-flexbox/
Flexbox is a CR, and it's starting to get pretty well supported: http://caniuse.com/#feat=flexbox
Whoopee.
However, they don't abstract over tasks like layout--you're still stuck with hacky CSS tricks and semantics.
You're correct, but that hasn't stopped people from trying.
I agree with your frustration about this article being newsworthy (it's 2013! How are we still figuring out vertical centering?!), though I would argue that CSS is great for its original use case and certainly has its place today. Why is it bad that CSS is entrenched in the Web?
Like any development technology, CSS has a learning curve and has several shortcomings (like vertical positioning), but I think it's a fine tool for the job and struggle to come up with something better.
A scrollable table with fixed headers is something that Visual Basic supported in 1997.
Of course it wasn't a use case of interest to the mighty browser vendor cabal so it has never been considered a priority. It does seem like it will be possible soon though with position:sticky https://air.mozilla.org/intern-presentation-ford/
We should probably prepare some sacrifice to the browser vendor gods to thank them for this small morsel they have thrown us.
CSS did borrow concepts from TeX at the time. I agree with you that if it had borrowed more from TeX maybe we would have avoided Javascript and also we would have benefited from an updated model for TeX and I said so at http://tex.stackexchange.com/a/6463/963.
Maybe is time to revisit and revamp both.
I have never seen this fully implemented except in the simplest of blog-engine layouts and toy examples. There is always some pollution of the markup, at the very least in the form of the ordering of elements.
A template-based approach would've provided the same semantics a lot sooner than a decade of half-baked CSS hacks.
Can you please elaborate how tables were the understanding of layout models 15 years ago?
The web was not thought to be interactive at that era, but to be printed, it's not the fault of the old specs that the web progressed significantly fast.
(1) The "H" in HTML stands for hypertext. (2) When has a web page ever printed well?
I said the web was tought to be printed, all the effort of CSS (circa 98) was about styling the web in a way that made sense for print, not for interaction.
In addition, if the web was truly meant to be printed — an assertion I have never heard before; it sounds blatantly false — then not adopting the massively successful TeX system sounds even more insane.
CSS makes life much, much easier.
That said, I use LESS. But while everything compiles down to CSS, it doesn't impact me much.