>This suggests a misunderstanding of the Standard Ebooks process, which allows continual incremental corrections to the authoritative source of individual books (in XHTML, on GitHub). So, a truly unique identifier would only be valid to the production output(s) from a particular state of the Git-repo sources.
Well, one of us has a misunderstanding. Just because the printer strikes off the printing number from the colophon for each subsequent printing, they don't actually issue a new ISBN. That stays the same. If they wanted to also include a version number too, I wouldn't mind that as well, but it's not nearly as necessary as this. I use the year as a rough version number in the file names as well.
>Recall also that final user content is made available in multiple formats, currently at least six. Example:
I don't need them to issue a number per file format, but if they want to... that doesn't bother me. That's sort of self-evident which of the formats it is, after all.
>I'd suggest the git-commit date, along with short substrings of author name and title, would suffice.
It doesn't. A number of authors have at one time or another have released books with similar or identical titles that are not the same book. This is the trouble... someone who uses or would use the books is asking for something that is missing but easy to supply, and instead of a "well gee, we never considered that, let us think about it" I have a dozen assholes crawling out of the woodwork to say "no, you're doing it wrong".
I need unique identifiers that are human readable. I just do. The world discovered this need for books before you were born. They invented a global standard, even. There is an entire field of science out there about this, that you seem to be ignorant of even existing. I've been doing this for years, and I keep bumping up against it. But you think it can be solved because you used git and know about hashes or whatever, and it's just like what you deal with in your software development job!