by Michael S. Kaplan, published on 2012/07/16 07:01 -04:00, original URI: http://blogs.msdn.com/b/michkap/archive/2012/07/16/10330046.aspx
Just yesterday, I was asked:
I'm running the Release Preview, and I found something confusing.
*SP* has weight, *ZWSP* has weight, and *NBSP* has weight, but *ZWNBSP* has no weight!
Is this a bug? Is it too late to fix before Windows 8 comes out?
Funny he should say that!
This was actually an intentional change that happened in the new version sorting in Windows 8, as of just after the Windows 8 Consumer Preview and a little time before the Windows 8 Release Preview.
This change directly undoes the change described in If a bunch of specific Unicode characters can no longer live in the same apartment together, can they really claim that they needed their space?.
It works like this:
The space character, U+0020, is given weight in the collation table.
For that matter, so do U+205f (MEDIUM MATHEMATICAL SPACE), U+3000 (IDEOGRAPHIC SPACE), U+1680 (OGHAM SPACE MARK), U+2002 (EN SPACE), U+2003 (EM SPACE), U+2004 (THREE-PER-EM SPACE), U+2005 (FOUR-PER-EM SPACE), U+2006 (SIX-PER-EM SPACE), U+2007 (FIGURE SPACE), U+2008 (PUNCTUATION SPACE), U+2009 (THIN SPACE), and U+200a (HAIR SPACE).
Perhaps more to the point, U+200b (ZERO WIDTH SPACE) and U+00a0 (NO-BREAK SPACE) have weight.
And for the last four versions of Windows:
and up to and including the Windows 8 Developer Preview and the Windows 8 Consumer Preview, U+feff (ZERO WIDTH NO-BREAK SPACE) has had weight too.
With a new major sorting version starting the Windows 8 Release Preview, this single inconsistency has been added once again, and technically made inconsistent....
As a bonus, we are also once again inconsistent with the Unicode Collation Algorithm, but I've been telling people for years that Microsoft does not use the Unicode Collation Algorithm.
If you want to make the other spaces also have no weight, then as I pointed out in I need my SPACE, symbolically speaking, the weight these characters are given is in the symbol range. So if you truly want to ignore the others, you can just call CompareString or CompareStringEx with the NORM_IGNORESYMBOLS flag. And you can go from there....
Now of course some people will cry out that U+feff is not just a space like the others, it is also the BYTE ORDER MARK (ref: Every character has a story #4: U+feff (alternate title: UTF-8 is the BOM, dude!)). Now in my humble opinion (many years and many building away from the group that owns the tables), the BOM has an even clearer semantic meaning attached to it, so ignoring it completely would not really be a linguistic or even a semantic requirement like some other characters like the ones I mention in Every character has a story #23: U+00ad (SOFT HYPHEN) and You've got to be kashidding me....
But the theoretical viewpoint that drove this change all those years ago was overruled by several important real world customer and partner scenarios.
And the "standards conformance" argument seldom holds water when we don't conform to that standard!
Second runner- up for this blog:
Reality trumps theory (almost always!)
Plus I wasn't there anymore, and neither was the linguist I was working with back then.
First runner-up for this blog, in case the title is unable to fulfill it's duties:
The people who are there trump the people who used to be there (most of the time!)
So anyway, if you see a ZERO WIDTH NO-BREAK SPACE in the Windows 8 Release Preview, don't be insensitive and comment it hasn't been eating enough lately!
skSdnW on 16 Jul 2012 9:04 AM:
The last four? Is 2012 a typo? If not, how can you leave out 2008R2 but not 2008? It is just Vista SP1 (Would you dare this kind of change in a SP?). Or is the list based on time and not version?
Mike Dimmick on 16 Jul 2012 9:30 AM:
@WndSks: I'm pretty sure Mike meant 2008 R2 in that list, not 2012.
2008 R2 is not just a service pack for Windows Vista. It is actually a whole new operating system - based on having a new kernel, shared with Windows 7, with some pretty fundamental modifications to the synchronization techniques protecting core system data structures in order to allow more than 64 processor cores/hardware threads. It did not warrant an R2 designation, it warranted a major version number; it should have a support lifecycle corresponding to a major release, not expiring at the same time as 2008!
skSdnW on 16 Jul 2012 9:37 AM:
2008 (non-R2) is the same kernel as Vista SP1 IIRC
Michael S. Kaplan on 16 Jul 2012 8:10 PM:
Yes, I meant Server 2008 R2, sorry about that!
go to newer or older post, or back to index or month or day