EPUB isn't all that different from what you're describing. It's bundled as a ZIP archive with a couple of XML metadata files - and the content is split into one HTML file per chapter or section to make it easier to handle - but the idea is the same.
There’s also an epub-namespaced set of attributes which extend XHTML with ebook specific semantics. But those typically aren’t necessary for the visual representation of books.
Hey, ChatGPT, tell me what's wrong with this person's comment.
> [T]he third comment violates the Cooperative Principle, specifically Grice’s Maxims of Relation and Manner, and ends up implying ignorance where there is none. Let’s break it down a bit more with that framework in mind:
> VIOLATION OF GRICE’S MAXIMS
> The second ["EPUB folks have continued to evolve their bespoke format instead of ditching it for something that legacy browsers already know how to handle"] commenter criticizes EPUB for continuing to evolve a packaging format that is not browser-native. They're not confused about what EPUB is—they're lamenting that it isn’t something simpler, like a plain web bundle a browser could just open.
> The third commenter responds by explaining what EPUB is, as if that somehow rebuts the original critique.
> Factually true.
> Entirely irrelevant in context.
> This failure to meet the relevance standard creates an implicature: the previous commenter must not have understood the format they were critiquing.
> THE IMPLICATURE TRAPS THE THIRD COMMENTER
> By stating something the second commenter obviously already knows, the third commenter unintentionally shifts the conversational footing in a way that belittles rather than builds. That’s why the tone feels off: not because of overt rudeness, but because the presupposition of ignorance is baked into the structure of the reply.
> FINAL THOUGHT
> The third comment reads like an attempted “correction,” but since the original comment didn’t contain a factual error, only a value judgment or proposal, this “correction” becomes a non sequitur—one that subtly undermines the prior speaker’s credibility while failing to address their actual point. That’s what makes it rhetorically broken, even if factually fine.
This is interesting, what kind of prompt or preprompting do you use to get it to notice things like Grice's Maxims violations? Mine is only set up to notice some informal fallacy violations
