Deprecate html4?

[silby]

T.P.W.HTML checks if it's in HTML5 mode in a dozen-odd places. This isn't fantastic and it got more intrusive as recently as last year via #9905, that jgm acknowledged as "academic".

I doubt that we're diligently checking up on the output of the html4 writer to see if it's still compatible with ancient browsers. Even browsers with deliberate limitations (e.g. Lynx, Dillo) support HTML5/the living standard to some extent, so -t html4 isn't necessarily helping to serve that audience.

And note that html4 is actually documented to produce XHTML 1.0 transitional; I'd be a little surprised if we're reliably producing valid XML. The way raw HTML is handled in Markdown means that test/writer.html4 doesn't parse as XML:

pandoc$ xmllint test/writer.html4
test/writer.html4:777: parser error : Opening and ending tag mismatch: hr line 600 and body
</body>
       ^
test/writer.html4:778: parser error : Opening and ending tag mismatch: hr line 595 and html
</html>
       ^
test/writer.html4:779: parser error : Premature end of data in tag hr line 592

^

So, I doubt we're serving very many people by maintaining the html4 writer, and if anyone is depending on it I doubt we are serving them particularly well.

[jgm]

We do aim to produce xhtml (also with html5 - polyglot html). But the writer tests include tests of writing raw HTML, which is just passed through -- including <hr> -- and that is why you get these issues in the test file.

Note that HTML4 is also used for EPUB v2, so this would require deprecating that format. But EPUB v2 is still used, particularly for older e-readers.

Hence I don't think the time is right to deprecate html4, but comments welcome.