Re: Conversion filters

["Robert Holmgren" holmgren@xxxxxxxx]

** Reply to message from Patricia M Godfrey  on Wed, 9 Oct
2002 19:28:13 -0400


> Is there some other listing
> someplace? Or did you change them for U2? Or does it just depend on the
> code page?

There's a listing in U2, largely derived from CHARSET. CHARSET's verbal
descriptions of what to expect from a character number are based on CodePage
437 (US-Ascii). Whereas almost everyone -- including the Word-4-Word filters
-- actually uses and expects 850 (Latin-1).

Here's what's happening:

The "extended [Speedo] characters" in range 782-909 have a tight relationship
with the characters in range 128-255.

When you are using CodePage 437, and you get line drawing characters in
128-255, 782-909 offer you access to the characters that you *would have
gotten* had you been using CodePage 850! Note that character 836 exactly
corresponds to character 182 (836-782 = 182-128 = 55th character in each
range). There are NO exceptions.

When you are using CodePage 850, these two ranges are exact duplicates. Now,
because W4W assumes 850, and therefore assumes them to be duplicates, it also
assumes you will use the 1-byte character rather than the 3-byte character
(*all* characters above 255 in the "extended set" are 3-byte characters). Now
rationale people are only going to avail of the extended set if it offers
something they can't get from the ordinary set (available on any computer), so
W4W feels entitled (reasonably, if you think about it) to ignore 782-909
altogether, all the time.

Note that the obverse is not true: if using 850, you do not get access to
437's line drawing and other US-Ascii characters in 782-909 -- they remain as
they are (850-ish) in *both* Codepages.

So if you're using 850, and you want to know what's on character 182 in
consultation with CHARSET, you have to perform a little mental math:
182-128=54; 782+54=836 -- look at the description on 836 to find out what's
actually on 182! What could be simpler? At one time, I had a bracketed number
after each extended number in range 782-909, e.g. "836 [182]" -- but nobody
knew what it meant, so I dumped it. Maybe I'll restore it (in TABLE
SPEEDOS). Let's see, where are those backups anyway... 1991...
February...

-----------------------------
Robert Holmgren
holmgren@xxxxxxxx
-----------------------------