`limitations' of OzTeX (was: fontinst with 8y.etx)
Wed, 17 Jun 1998 12:44:03 -0700 (PDT)
Using data from my own tools, I wrote:
>> My own custom encodings, which are based on the PDFDocEncoding
>> wouldn't work particulally well on these Windows DVI previewers. If we
>> rate an encoding's compatibility with Windows ANSI as N+M, where N is
>> the number of slot clashes, and M is the number of glyphs that map to
>> empty slots in Windows ANSI (and thus lower numbers are better), we get:
>> TeXBase1Encoding (aka 8r): 4+21
>> TeXnANSIEncoding (aka LY1/8y): 7+35
>> PDFDocEncoding: 23+16
>> my current custom encoding: 28+26
>> ECEncoding (roughly T1): 63+41
>> Thus we can see that 8r is most conciliatory towards these Windows
... leading Berthold to reply:
> I am not sure what those numbers mean, since the obvious interpretations
> leads to some contradictions :-)
> First of all, 8r and 8y have the same overall glyph complement, so the
> numbers must be wrong. Each includes the 15 glyphs missing from
> Windows ANSI yet found in the standard 228 of typical text fonts.
> Each also includes ff, ffi, ffl, dotlessj. So the +21 for 8r and +35 for 8y
> should be BOTH +19 by my counting. I wonder whether you have
> added into the 8y total the glyphs that are repeated for convenience?
8r and 8y *don't* have the same glyph complement 8y includes `cwm',
`nbspace' and `sfthyphen', which are not in 8r. Also, I suppose I
shouldn't have said ``the number of glyphs that map to empty slots''
and instead said ``the number of empty slots into which glyphs are
> As for conflicts, both 8r and 8y take the `standard' departure from
> Latin 1 (and hence Windows ANSI) by replacing quotesingle in 39
> with quoteright, and replacing grave accent with quoteleft in 92.
> They also both have asciicircum instead of circumflex accent in 94,
> and asciitilde instead of tilde accent in 126. So I count 4 conflicts
> with Windows ANSI in *both* 8r and 8y.
Perhaps my tools were overly zealous in their reporting here. This
problem again stems from those three glyphs above. There seems to be
disagreement on what the three glyphs that 8y calls `cwm', `nbspace'
and `sfthyphen' should be called. Adobe refers to `nbspace' as
`nobreakspace' (e.g. in Adobe's chsttabl.pdf) but usually replaces it
with space in encodings, `sfthyphen' is sometimes called `softhyphen'
(e.g. on the HP LaserJet 4000) and `cwm'(*) is sometimes called
`compwordmark' (although I can't remember where right now).
* In fact, I've never seen a `compwordmark'/`cwm' glyph in any
font, so I have no idea what it might look like, or whether it
really needs to be in 8y.
Another issue is that I don't have a definative source for the Windows
ANSI Encoding. One source is WinLatin1Encoding present in my HP LaserJet
4000, and another is the WinAnsiEncoding described in the PDF 1.1
specification -- and neither has sfthyphen, nbspace or cwm.
Given the amount of naming diagreement, I wonder if there is any good
reason to include these glyphs in 8y.
P.S. I've enclosed the output of my encoding comparison tool, so you can
see the kind of data tool I was working with.
unix% compare-encodings 8r.enc texnansi.enc
TeXBase1Encoding vs. TeXnANSIEncoding
232 glyphs common to both encodings
19 slot clashes
210 slot matches
21 empty slots in TeXBase1Encoding filled in TeXnANSIEncoding
4 filled slots in TeXBase1Encoding empty in TeXnANSIEncoding
3 glyphs only in TeXnANSIEncoding
cwm nbspace sfthyphen
Slot TeXBase1Encoding TeXnANSIEncoding
05 hungarumlaut dotaccent
06 Lslash hungarumlaut
07 lslash ogonek
08 ogonek fl
0B breve ff
0C minus fi
0E Zcaron ffi
0F zcaron ffl
10 caron dotlessi
11 dotlessi dotlessj
12 dotlessj grave
13 ff acute
14 ffi caron
15 ffl breve
1E grave OE
1F quotesingle Oslash
5E asciicircum circumflex
7E asciitilde tilde
AD hyphen sfthyphen