Re: [CSS21] response to issue 115 (and 44)

> Well, in my reality there are browsers which behave rather different
> from what you describe, Internet Explorer for Windows for example, e.g.

On Windows IE 5.5, auto-detection has been broken for as long as I've
known it; that suggests that its not used by enough people to be a
marketing priority (if you print with it enabled the pages are blank
except for the headers and footers).

The reality, with modern browsers, is that after the first page
that is gibberish because the browser defaulted to iso-8859-1, the
user goes to the View | Encoding menu and selects their local
chracter set.  As the page didn't have a character set, this sets
the browser's default character set[1], which is the only heuristic
used by browsers in normal use.

As an added complication, at least outside the primary areas for CJK
character sets (and probably for historic reasons, even within it)
people often use add on software, for Chinese, e.g. Richwin, NJStar,
etc. which intercept display requests for Windows 1252 and
generate characters, as though the byte stream had really been a
big5 or gb2312 byte stream[2].  In that case, the browser may well
be set for iso-8859-1.  Modern web browsers are exceptional on
Windows 98, ME, etc., in that they are Unicode aware.

I know of one person who is actually thinking of downgrading from XP to
98, in order to use such software, because, for some reason, they
can't get the XP software to work in Chinese (possibly because they
are using AOL for internet access).

I'm afraid the real world situation on i18n on the web is messy because
early browsers didn't consider i18n a priority, but non-US users
did consider being on the web a priority.  There are a lot of workaround
out there of which iso-8859/1 or windows 1252 users aren't really aware.

[1] This menu is generally multi function, being a character set status
indicator and a temporary override when the page does claim a character
set and a permanent override for all such pages, when the page doesn't.

[2] It is possible that some of these packages now have Unicode awareness,
but they do tend to come with their own, non-Unicode, fonts.

Received on Saturday, 21 February 2004 14:45:07 UTC