W3C home > Mailing lists > Public > www-style@w3.org > May 2010

Re: [css3-lists] Proposal for a generic numeric list-style-type

From: Aryeh Gregor <Simetrical+w3c@gmail.com>
Date: Mon, 17 May 2010 14:57:33 -0400
Message-ID: <AANLkTilr9jg1Eajul7b1uY-bu7YiB5T24wp3WZ1dsdgU@mail.gmail.com>
To: Markus Ernst <derernst@gmx.ch>, Gabriele Romanato <gabriele.romanato@gmail.com>
Cc: www-style@w3.org
On Mon, May 17, 2010 at 9:15 AM, Markus Ernst <derernst@gmx.ch> wrote:
> Now usually the type of numeric list markers depends on the language.

Sometimes, yes.  Often, it seems, not.  Empirically, the Hebrew,
Arabic, Chinese, and Japanese Wikipedias all use decimal numbers for
<ol>, even though they all have native numbering systems too supported
in CSS, and there are certainly admins on all of those wikis who would
know how to change it.  The Persian Wikipedia uses list-style-type:
persian for <ol>, however.  Do any browsers ship different
list-style-types for different language editions?  Any other data we

I suspect it's not going to be so straightforward to figure out which
languages need native list markers and which don't.  My experience
suggests that at least for some languages, whatever you choose, some
users will think native numerals feel weird ("All other websites use
Arabic numerals, it feels wrong for a website to use native") and some
will think Arabic numerals are wrong ("All books use native numerals,
it's completely standard").

My guess is that this isn't something we want hardcoded in a spec.
Individual websites will have to listen to their specific users.  It's
really not such a big authoring burden.

On Mon, May 17, 2010 at 11:13 AM, Gabriele Romanato
<gabriele.romanato@gmail.com> wrote:
> Interesting. Honestly, I think this would be rather expensive in terms of
> CSS rendering, though the language information is something that a UA should
> know well, provided that a web document contains the aforementioned language
> information. Otherwise, the UA should use its default language information
> defined by the UA user. The problems is that a UA should perform a massive
> comparison between a list of supoported languages and the corresponding list
> of list-style-type values for a given language X.  What is the impact on
> performance?

I'd think not much.  It should be no worse than having a list of
selectors like ol:lang(fa) { list-style-type: persian }, although it
might be harder to implement.  "Massive comparisons" like this are
awfully efficient using things like, say, hash tables -- the
comparison itself would be negligible compared to the type it would
take to lay out, I imagine.  (But current implementations might not
have the language info handy at this point in the code, I don't know.)

> Further, CSS specs provide yet a basic support for
> list-style-type internazionalization, though it could be extended, as you
> say. However, there's also the problem of rendering a style information in
> the form of an alphabetic system after a given number of letters. For
> example, in Western languages we have 27 letters in our alphabets. Specs say
> that after the 27th letter, the rendering of list-style-type is undefined.
> What happens to a list of 28 items? So I think that what you're proposing
> right now is not so feasible, though interesting.

The spec defines 26 values for {lower,upper}-latin, not 27.  The
behavior after the 26th value is well-defined: it goes z, aa, ab, ac,
..., az, ba, ... ad infinitum.  See the spec:
Received on Monday, 17 May 2010 18:58:10 UTC

This archive was generated by hypermail 2.4.0 : Friday, 25 March 2022 10:07:46 UTC