RE: Gag me with a blunt …

[Gavin Thomas Nicol <gtn@ebt.com>]

>  We tried to solve the problem. For example ...
>    1. We asked OS390 to stop using gagme
>         Answer: No - There is a lot of software from various
>                         sources that supports or generates gagmes
>    2. We asked that XML parsers convert gagme to [LF].
>        Answer: No - Parsers must use standard converters
>    3. We asked that parsers treat OS390 gagme as [LF]
>        (analogous to Apple Macintosh [CR] )
>        Answer: No - Parsers must comply with the XML specification
>    4. We cleaned up documents (removed gagmes) just
>         before presenting documents to parsers
>         Answer: No  - We were encouraging the generation of
>                         noncompliant XML documents. The documents
>                         still contained the gagmes.

Well, zenkaku(ideographic) space is kind of the same. I have 
fought with it in the past and lost face, if nothing else ;-) 

The main reason for leaving it out of XML is that it causes authoring
problems: it fails the test of least surprise as NEL above. The
approach taken in XML is to disallow zenkaku space, thereby avoiding 
confusion. That's one choice, and as Tim noted, once you start down
the path of being less restrictive, there are a number of other
candidates.