This topic is a side product of my experiments with a Typst-based ebook reader demo (described elsewhere on this forum), but I think it is worth isolating this topic from the discussion of any particular app. I will also leave aside the use case of “researchy” books here; it is super interesting, but this topic is about a different thing. So:
It has long been noticed that ebooks as we have them compete devastatingly well against cheap paperbacks but do not have much to offer in the “fine press” segment. Nobody seems to be producing ebooks in the way Folio Society makes traditional hardbacks; I, for one, often feel admiration towards the craft that went into a paper book, but never had this experience with an ebook.
When I started my own limited experiments with Typst-based ebooks, I expected the new possibilities to be mainly about well-executed formatting and nice page layouts, all taken seriously rather than as optional (like e-readers treat css). I quickly realized there was another, arguably bigger advantage: the full, real support of conditionals, that is, of the good old if-then-else with potentially complex expressions involved. Especially the possibility to build conditions involving both internal content and externally provided formatting parameters seems to mean that Typst is offering more expressive power here than e.g. CSS media queries (even if e-readers observed the latter faithfully).
A lot of exciting formatting decisions (cleverly positioned illustrations, exuberant initials, unusual headers or footers) have limitations in terms of page proportions, or page size, or both. With EPUB, we learned to just avoid them. With Typst, we technically can ensure they work for widespread screen proportions (maybe in three different ways for three different buckets of page proportions), and provide a fallback for whatever remains.
The same illustration can be embedded into the text on a square page differently from what it does on a tall and narrow page, or sometimes omitted altogether; a header can be produced differently for a two-page view and for a one-page view; an initial replaced with plain text if it cannot be positioned nicely. That would, obviously, require an agreed-upon, carefully defined set of formatting parameters passed from the e-reader or converter to the Typst compiler, but that seems like a solvable problem.
Interestingly, when targeting the “fine-press” segment, “slow” stops being a bad word; compiling your new edition of Pride and Prejudice into a (for example) pdf with the parameters of your favorite e-reader is potentially a pleasurable ritual, not an annoying delay. As for producing such editions, I can easily imagine embracing an artisan hobby like this if there were infrastructure for it.
An ebook that does not stretch like rubber in all directions but rather rethinks itself depending on the constraints provided, showcasing the foresight and craftsmanship of the bookmaker: this is something Typst appears to support already.
Putting this into practice will not be frictionless of course; to highlight one particularly challenging bit, a complex enough book will need some sort of “unit test coverage” of various compilation setups to check that it is working as intended. But still it seems that, bottom line, things are becoming possible that were not possible before.
Has anyone experimented with something like this before, or else, does anyone feel inspired to?