I think this kinda isn't what people mean when they say end-user programming, but I think it counts: I found it interesting how the structure of the entire system being accessible and editable by the end user at runtime, including the core algorithms, is what gives the system it's power It seems some problems can be only solved in particular cases (Incidentally, if anyone knows of any attempts at composable approaches to typesetting algorithms, I'd be interested, the problem is interesting as typesetting a document is essentially solving a global optimization problem, not a bunch of local ones, so decomposition is hard)

Whoa, I haven't seen a new typesetting system in years! Thanks for the share

Most people will just use Adobe InDesign (or Quark) for setting materials, which handle Japanese beautifully. These Linux/TeX systems have horrible user interfaces and are only for nerds. Nothing end-user friendly about them. Knuth was a maniac who refused to use the renderers available to him; he ignored the industry standards of Adobe Type1 and then later TrueType and invented his own. It is so painful to have multiple renderers and font formats. It is a tragedy of the computer industry that it can never agree on basic standards whether it be the punchcard dimensions, or the ASCII/EBCIDIC code, etc., These battles create chaos and confusion for the users. Just like the Beta/VHS battle, or the HD-DVD vs BluRay, it is always some lengthy never-give-an-inch industry competition that bruises both sides. Remember when DVD disks were either +R or -R? That meant which frequency laser beam was being licensed in the burner. Gad, that was so stupid. I wish TeX would just go away already. It was a bad design, but the mathematicians and academics have kept it alive all these years.

It seems to me that the problem with LaTeX is its lack of modularity: there is no typesetting system separate from the TeX language, even though the syntax really seems like a separate concern. I would like to see technologies like TeX and HTML/CSS decomposed so that the layout functionality would be separate from the textual language(s), separate from the primitive rendering system, separate from JavaScript. And then I'd like to see layout decomposed further into modules like, for HTML, "normal", "flexbox", "table" and "grid"... again and again everybody makes the mistake of making a monolithic system without clear boundaries. It can be good for adoption when there's no good competitors, but in the long run it limits innovation. (The web is kind of oddball in being separated into HTML/CSS/JS but treated like a monolith at the same time, so sure your HTML can always contain CSS and JS code... you've got the drawback of learning three languages but without much of the benefit of modularity - edit: okay, there are real benefits of having a monolith, but none that couldn't also be found in a more modular architecture that is composed in a standard way across browsers)