Not only that, it's based on byte offset and length: "SourceContent=8BIT,1279,492,https://archive.org/download/MoeJusteOrigins/8-BigBang-LostS... "? Is the idea that the target file is never modified (or any new versions get a new URL)?

Also, I think that HTTP got it right when encoding is returned with the document (rather than specified in the link), but if you've already made the mistake of dealing with byte offsets you're tied to a particular representation.