On Thu, Nov 13, 2014 at 12:56 PM, Evan Laforge wrote:

On Wed, Nov 12, 2014 at 2:44 PM, Wojtek Narczyński wrote:

...

How to create a DSL for HTML generation, that would (statically) allow only valid HTML to be generated? Parent - child relations, valid attributes only.

Or does such a DSL already exist, perhaps?

WASH did that, a long time ago. They mentioned that they intentionally relaxed the rules, since being totally correct was annoying to actually use. There are docs here: http://www2.informatik.uni-freiburg.de/~thiemann/WASH/

Well, a lot of folks felt that any validation beyond "it works in my browser" was annoying to actually use. I, on the other hand, used NSGML in emacs and the actual DTD without much problem. But there were multiple DTDs, some of which also relaxed things because so many people felt the strict rules were annoying to actually use. It didn't seem to catch on, so maybe it was too heavy for its benefits.

Based on a decade of trying to get people to write valid HTML - so that it would work in browsers other than the most popular two or three on the most popular two or three platforms - most people don't care about "correct".. I've had people - including employees of the FSF, who I would expect otherwise from - tell me that their HTML was tested on some limited set of browser/platform pairs when I pointed out it was incorrect and hence broke on some obscure browser/platform combination. Or even on one of those browser/platform pairs with a configuration change. This also seems to be true in the Haskell community. The web platforms help catching the obvious problems you find testing against a browser. But - like most popular web platforms - break the tools that help with the creation of valid HTML Validating HTML against a DTD is much easier than doing so against browsers, but my clients never did it unless I set it up for them. Doing so helped find more than one really obscure bug, and on one occasion kept my clients web site working when a major browser tweaked how it handled broken code and broke a lot of web sites. I think this may be related to why ADTs are a hard sell. While they help a lot writing correct programs, but for just writing working programs not so much.