From: Asmus Freytag (asmusf@ix.netcom.com)
Date: Thu May 22 2008 - 12:37:04 CDT
On 5/22/2008 9:59 AM, John Hudson wrote:
> David Starner wrote:
>
>> On Wed, May 21, 2008 at 10:43 PM, John Hudson <john@tiro.ca> wrote:
>>> The key word here should be *glyph*. Correct cultural norms for spacing
>>> punctuation should not be a text encoding issue at all, any more than
>>> spacing any other glyphs should be an encoding issue. These should be
>>> display issues, handled via font intelligence and language tagging.
>
>> Taken most literally, that's obviously not a common practice at all; I
>> note the spaces after your commas and periods, and the examples I've
>> seen without them have struck me as erroneous and deficient.
>
> A word space is a semantic separator. The space that French
> typographers traditionally place before some punctuation marks is not.
> It doesn't change the meaning of the text whether that space is
> present or not.
It may not technically change the meaning, but if you consider
robustness of text interchange, you also need to consider how acceptable
a fallback is to the average reader. I cannot gauge this reliably for
this issue, but from what I read it seems that this is at least a
borderline case, in the sense that from what people write, too much
space (a full space) is a better fallback than none. If that's the case,
then letting the user key in the encoding for the fallback presentation
would be advantageous.
Finally, there's the issue of whether automatic layout software can
reliably make the determination when to apply French conventions.
David's example strikes me as one where it's more than possible to have
documents for which the style guides (whether formal or informal) do not
call for French punctuation rules to be applied in the case of short
inserts. Yet for many other purposes (spell checking) one wouldn't want
to tag the insert with an incorrect language.
It seems to me a general principle that the more visually obvious some
layout rule turns out to be, the more you need the user in control about
its invocation, and the more damage could be caused by relying solely on
implicit models.
A./
This archive was generated by hypermail 2.1.5 : Thu May 22 2008 - 12:40:15 CDT