html5 option was added in 2010 and marked as "not experimental" in 2011
but is not the default.
According to http://caniuse.com/#feat=html5semantic, current versions of
all recent versions of all major browsers - even IE (9+) - support the HTML5
semantic elements (
<section> etc.), except for
<main> which IkiWiki
doesn't use anyway. However, IE 8 is not a current version, but has ~ 4%
market share and doesn't support
<section> and friends; so there's still
a compatibility concern there.
In particular, non-HTML5 mode uses
<!DOCTYPE html PUBLIC
"-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
which doesn't allow newer markup like the
role attribute, so we can't close
add aria landmarks to make ikiwiki websites more accessible while
remaining XHTML 1.0 Strict. The recommended pseudo-doctype for HTML5, and for
HTML with ARIA markup, is
(I do think we should continue to use
<xml-compatible-tags /> and output
well-formed XML so people who want to do XSLT tricks with IkiWiki's output
can do so, though.)
In practice, real browsers have never actually implemented a strict XHTML mode:
they've always parsed
text/html as "tag soup", because they need a tag-soup
parser anyway, and nobody wants to maintain two parsers.
Kai also wants a HTML5 doctype for more mobile friendly default themes.
set html5 to 1 by default but retain the dual-mode templates, breaking IE 8 by default
remove the option and always behave as if it had been 1, simplifying the templates and breaking IE 8 unconditionally
either of the above and include html5shiv to de-break IE 8
change the doctype to
<!DOCTYPE html>unconditionally, stop trying to limit ourselves to XHTML 1.0 Strict (use HTML5 features that degrade gracefully, like ARIA roles), but avoid using the new elements like
<section>that require specific browser support unless
html5is set to 1. That would get rid of the backwards-compatibility concerns while keeping the ability to use post-2000 markup; we can use
html5to mean "be more enthusiastic about HTML5 features even if they might fail on older browsers".
Using the HTML5 doctype does mean we lose the ability to validate the output
against a DTD (as
wdg-html-validator does), but DTDs have very little to
do with practical browser compatibility in any case.
At the moment my preferred option is the last, for which see my
branch. I'll apply this at some point if there are no objections.