[pwdisswordfish8]

That doesn't change the thrust of the comment. If you're trying to work around spotty support and detecting the feature is "tricky", then change the program so it performs a small power-on self-test against a known dataset.

(And what's the point of crafting a comment in this tone? Is it supposed to be a retort? Whether or not Netlify is doing the wrong thing, if it works in Chrome, but not in Firefox, then that's a materially relevant fact. Don't rely on implementation details and a hope that they won't break in the future.)

[zinekeller]

Because both Chromium and Gecko follow the IETF (RFC)/W3C specs about this, what Netlify is doing is plain out-of-spec, so what Chromiun and Gecko are doing are implementation details that is explicitly marked as "okay, if you encounter a stupid server that is somehow explicitly advertising range-request support but does it incorrectly, you can do anything and you're still compliant". Drop the request (like Firefox)? Compliant. Silently trim (like Chrome)? Still compliant. Just give zeroes matching the announced length? Yes, still compliant even if you think that's stupid. If you didn't get this simple fact (that is easily verifiable by opening your favourite browser's devtools or even in Fiddler), I don't know how you're not getting this. Yes, it's using heuristics, but Netlify announces support for range-requests so there's no heuristics here to do, either Netlify must remove the header announcing support or Netlify fixes this problem before we talk about heuristics.

[domenicd]

A bit late for the original discussion, but in case it helps future readers, I've filed https://github.com/whatwg/fetch/issues/1295 to see if we can get the spec and browsers aligned. (My guess is we will update the spec to allow this and Firefox will update to align with the new spec, since the direction of specs over time is always toward more-lenient.)