From: Frank Ellermann (nobody@xyzzy.claranet.de)
Date: Tue Jan 23 2007 - 14:06:50 CST
Doug Ewell wrote:
> Before you give up for good, try reading the Appendix of UTN #14.
s/reading/porting/ - of course I've read it, but I intentionally
tried to implement it from scratch based on the spec. All these
windows are confusing...
Implementing BOCU-1 is straight forward based on the 2006-02-03
memo, still missing in my script is a decent error handling, see
<http://purl.net/xyzzy/src/bocu.cmd> - the BOCU-1 part is about
170 lines.
> SCSU is very stateful.
...indeed.
> I think you may be thinking of some other algorithm.
I'm talking about the algorithm with 256-34 possible lead bytes
from 0x22 up to 0xFE. There's no octet left to introduce more
than three trailing bytes. My script does not yet check that
the input (or output resp.) must be a valid code point, I've
done that only for the various UTFs (1,4,7,8,16,32).
> "Necessary to... the implementation of" means you cannot
> implement BOCU without infringing on IBM's patent
Maybe. It's legalese, it explicitly says "may". Maybe this
deponds on living in the U.S. Or maybe it depends on how much
you've annoyed IBM elsewhere. I'd guess that you're seriously
in trouble if you claim to have "invented" another profile of
BOCU (slightly different from BOCU-1).
> I tried for months to obtain a developer-friendly clarification
> of this restriction -- something akin to the "freely available"
> clause in UTR #16 (UTF-EBCDIC) -- and was utterly unable to do
> so.
That's bad. If I'd try this and all else fails I'd ask Brian
Mark, or Markus if they've an idea how that's supposed to work
within IBM.
Frank
This archive was generated by hypermail 2.1.5 : Tue Jan 23 2007 - 14:14:47 CST