On Wed, 7 Sep 2016, Thorsten Glaser wrote: > In real life, it does help: if none is specified, various user agents > may (and do¹) guess. This occasionally helps.
And occasionally leads to security vulnerabilities: http://nedbatchelder.com/blog/200704/xss_with_utf7.html Even aside from those kinds of issues, if the early web has taught us anything, it’s that it’s worse to serve pages that happen to work in some user agents than pages that fail predictably, because browser-specific mistakes are less likely to be discovered and fixed. > ① and do: unfortunately, Mozilla derivates always default to latin1 for > many UTF-8 files… Like I said, that’s the standard behavior for ‘text/plain’. To correctly serve UTF-8 plain text, you need to mark it ‘text/plain; charset=UTF-8’. Do you really want to serve pages that don’t work in major browsers, anyway? > I don’t want to play bts control pingpong with you, but please consider > this patch nevertheless, and feed it upstream. Sorry, my answer is still no. If you want upstream’s opinion (which I expect will be the same), see the directions in Documentation/SubmittingPatches. Anders
