path: root/host/resources/fonts/issues.txt

Design issued related to the -misc-fixed-*-iso10646-1 fonts
-----------------------------------------------------------

$Id: issues.txt,v 1.11 2006-01-05 20:31:45+00 mgk25 Rel $

This file contains various technical notes from people who have
contributed glyphs.

---------------------------------------------------------------------------
Here are some short notes on certain problematic glyphs that people
easily make wrong:

  U+0027 (APOSTROPHE)
    This should be a neutral (vertical) glyph, usually a single
    stroke version of U+0022 (QUOTATION MARK).

  U+0030 - U+0039
    Please make these the same height as capital letters, and make
    <zero> different from <capital-o> usually by making it narrower,
    or, perhaps by adding a diagonal stroke inside it.

  U+0042, U+0044, U+0045, U+0046 (B, D, E, F)
    Please lose the bogus pseudo-serifs in fonts that aren't otherwise
    serifed, especially in small fonts.

  U+004A (J)
    The top should look like :  ### , not ###
                                 #          #

  U+0060, U+00B4 (GRAVE ACCENT and ACUTE ACCENT)
    These should be mirrored versions of each other.

  U+0061 (a)
    Be careful if you make this is cursive a. See notes on U+0251.

  (accented capitals)
    You may have to make the capitals smaller for this to work. Do so.
    Leave a gap between the accent and the capital, unless this would
    make the capital the same height as regular small letters, which must
    be avoided if all possible.

  U+00DF (sharp s)
    This is NOT a <beta>. It is not supposed to look like one. It looks
    more like a ligature of a <long-s> (U+017f) followed by a
    <round-s> (U+0073) with a line linking the tops of them.

  U+010f (lowercase d with caron)
    This is potentially ugly. Feel free to reduce the height of the <d> if
    needed.

  U+0123 (lowercase g with cedilla)
    Don't bother drawing a cedilla below, as the tail of the g would
    interfere. Instead, follow the <lithuanian?> convention, and place a
    turned comma above (U+0312).

  U+0145, U+0146 (n with cedilla)
    It is OK for the cedilla not to be attached to the letter!

  U+018D (small turned delta)
    ?What should this look like?

  U+0194 (capital gamma)
    ?What should this look like?

  U+01A2 (oi)
    ?What should this look like?

  U+01A5 (oi)
    ?What should this look like?

  U+01A9 (capital esh)
    Yes, this is a <Sigma>.

  U+01C3 (retroflex click)
   Try and differentiate from punctuation, by making the stroke
   thicker at the top.

  U+01C4, U+01C5 (dz with caron)
    If you need to shrink the capital, it is probably best to shrink all
    the capitals in both these glyphs.
  
  U+01BF (small wynn)
    Like a p, but with a diagonal line at the bottom of the loop?
    This used to be used to represent /w/ in English, but got abandoned
    due to confusion with <p> and <thorn>, so don't worry if it looks too
    much like p : history agrees with you. ;)  

  U+01E2, U+01E3 (ae with macron)
    The line should be above both the A and the E components.

  U+0222, U+0223 (ou)
    Like an 8, but with a broken top. In reality, it is a ligature of
    <Omicron> and <Upsilon>, so if you have enough pixels, try making 
    it look like that.

  U+0251, U+0252 (script a, turned script a)
    If the default <a> in your font has is a <script-a>, try to make this
    an exaggerated <script-a>. If <alpha> is not the same as <a>, try
    using the same glyph as that.

  U+0253 (b with hook)
    This should look like a regular b, but with a hook from the left stroke,
    extending for maybe 80% of the width of the letter.

  U+025F (small letter turned f)
    The hook of the inverted-f should be below the base-line, and the
    highpart of the glyph should be at x-height.
    
    Note : this is listed in "Phonetic Symbol Guide", as being a barred
    dotless j.

  U+0260 (g with hook)
    This should be a modified U+0261, not a modified <g>, which might have
    a loop below.

  U+0264 (rams horn)
    This needs to be graphically distinct from <gamma>, and <ipa-gamma>.
    Emphasize the horns. It is normal character height.
  
  U+0265 (small letter turned h)
    ! This needs investigating !
  
  U+0278 (small letter phi)
    No superflous serfis, please.

  U+027[ABCD], Where, wrt baseline?
  
         (turned r with long leg,
	  turned r with hook,
	  r with long leg,
	  r with tail)

    Follow 9x18. It is right.

  U+0284 (small letter dotless j with stroke and hook?)
    See 9x18. Yes, it's an esh with a line across near the bottom of the
    vertical.

  U+0288 (t with retroflex hook)
    Extends below baseline.

  U+028B (v with hook)
    The closest leter to this is called "SCRIPT V" in PHONETIC SYMBOL GUIDE.
    See 9x18.

  U+0299 (small capital b)
    Lose the serifs.
  
  U+0283 (turned y)
    Above x-line.
  
  U+029E (small letter turned k)
    Goes below the baseline.

  U+03C6, U+03D5 (GREEK SMALL LETTER PHI and GREEK PHI SYMBOL)
    Note that the example glyphs for these two were accidentally
    swapped in Unicode 2.0 and ISO 10646-1:1993.

  U+22C0 .. U+22C3 (n-ary and/or/intersection/union)
    These should just be larger versions of U+2227 .. U+222B,
    same size as n-ary sum (U+2211) and product (U+220F). The
    bold glyphs in Unicode 2.0 are bad, the glyphs in
    ISO 10646-1:1993 are fine.

  U+2308 ..U+2305 (floor and ceiling)
    These should be like square brackets with the top or bottom
    bar missing. (Rounding operators, invented by Iversion for APL)

  U+2400 .. U+2424 (ASCII control code pictures)
    The letters should be arranged diagonally falling like in
    ISO 10646-1:1993 and not on a horizontal line like in Unicode 2.0.

---------------------------------------------------------------------------
A note by Markus Kuhn on quotation marks and grave/acute accents
(1999-07-16):

The old misc-fixed-* fonts had the characters

  U+0027  '  APOSTROPHE

and

  U+0060  `  GRAVE ACCENT

shaped as mirror images of each other, such that they could also be
(ab)used as single opening and closing quotation marks. This was
probably influenced by how TeX uses these characters and sanctioned by
very early versions of ASCII, but it conflicts with many other
well-established conventions, namely

  - the requirement that U+0060 GRAVE ACCENT and U+00B4 ACUTE ACCENT
    logically have to be mirrored versions of each other and that they
    both should look like accents (straight lines) and not like curly
    quotation marks
  - how these characters appear in the ISO 646, 8859, 10646, etc. standards
  - the Unicode 2.1 requirement that U+0027 be a "neutral (vertical) glyph
    having mixed usage"
  - the way these characters are commonly depicted on keyboards
  - the way these characters appear in many other commercial Unicode fonts
  - the fact that Unicode provides two other characters, namely

      U+2018  LEFT SINGLE QUOTATION MARK
      U+2019  RIGHT SINGLE QUOTATION MARK

    in order to provide the directional curly quotation marks and also
    the curly apostrophe that TeX users are used to enter with ` and '
  - the fact that U+2018 and U+2019 are in practice already very
    widely used for these purposes (e.g., by Microsoft Word)
  - the fact that the semantics of U+0027 corresponds to the
    vertical apostrophe and undirected quotation mark found on
    old typewriters
  - the fact that Adobe officially maps Unicode to Postscript's accent,
    apostrophe and quotation characters as follows:

      U+0022 = quotedbl             QUOTATION MARK
      U+0027 = quotesingle          APOSTROPHE
      U+0060 = grave                GRAVE ACCENT
      U+00B4 = acute                ACUTE ACCENT
      U+2018 = quoteleft            LEFT SINGLE QUOTATION MARK
      U+2019 = quoteright           RIGHT SINGLE QUOTATION MARK
      U+201A = quotesinglbase       SINGLE LOW-9 QUOTATION MARK
      U+201B = quotereversed        SINGLE HIGH-REVERSED-9 QUOTATION MARK
      U+201C = quotedblleft         LEFT DOUBLE QUOTATION MARK
      U+201D = quotedblright        RIGHT DOUBLE QUOTATION MARK
      U+201E = quotedblbase         DOUBLE LOW-9 QUOTATION MARK


Therefore, the shapes of the U+0027 and U+0060 characters have been
fixed in the X11 *-iso10646-1 font versions and differ from those of
the old Latin-1 versions of the same fonts. This will discourage
people from continued abuse of the GRAVE ACCENT character as a single
left quotation mark, which looks really horrible with many non-X11
fonts in use today. Please fix software that writes text such as
`quote' and better let it write 'quote' instead (or even use U+2018
and U+2019 if Unicode output is feasible).

References:

  - Michael Everson: On the apostrophe and quotation mark, with a note on
    Egyptian transliteration characters, Working Group Document
    ISO/IEC JTC1/SC2/WG2 N2043, 1999-07-24,
    <http://www.dkuug.dk/JTC1/SC2/WG2/docs/n2043.pdf>

  - http://partners.adobe.com/asn/developer/typeforum/unicodegn.html

---------------------------------------------------------------------------
From: Birger Langkjer <birger.langkjer@image.dk>
Date: Fri, 6 Aug 1999 15:21:55 +0200

About accents: We discussed it before and decided we didn't have to be
overly respectfull of the original font. I went down to the library and
borrowed some books in Polish and Turkish to look at accented
characters in their natural setting so to speak. As a result I moved
all the accents on lower case letters down a pixel so that they are
relative to the letter rather than on the same height. It really
looks a lot better now that I look at it again.

---------------------------------------------------------------------------

It is a good idea to have some references for various scripts.

International Phonetic Alphabet (IPA):

  http://www.arts.gla.ac.uk/IPA/fullchart.html

  A good book to read is :
  
   "Phonetic Symbol Guide", 2nd edition, by Geoffrey K. Pullum, and
   William A. Ladusaw, ISBN 0-226-68536-5. Much of the advice on IPA
   characters is derived from this.

Armenian:

  http://moon.yerphi.am/~hovik/Armenian/

Others?

New Unicode 3.0 characters are described in the various ISO 10646-1
(draft) amendments available on

  http://www.indigo.ie/egt/standards/iso10646/pdf/

Many people agree that the glyphs found in ISO 10646-1:1993 are better
and more typical for the represented scripts than thoise found in the
Unicode 2.0 book. If you have a change to get access to ISO
10646-1:1993, then use it.
---------------------------------------------------------------------------
Comments by Constantine Stathopoulos <cstath@irismedia.gr>
(1998-10-19):

I have made some changes from what would be considered as strictly
correct:

1) Capital combinations with psili+oxia, psili+varia, dasia+oxia and
dasia+varia (e.g. U+1F0A to U+1F0E) are definitely incorrect compared
to the uncombined/spacing diacritics (U+1FCD, U+1FCE, U+1FDD and
U+1FDE). That was necessary due to the 6x12 cell limitation, but is of
no consequence, since in such fonts accented capitals are typed as two
characters: spacing diacritic + unaccented capital letter.

2) Ypogegrammeni in combined small letters (e.g. U+1F87) is also
different from the uncombined/spacing ypogegrammeni (U+037A) due to
the matrix limitations. The resulting characters are not incorrect;
they are just different in style, but completely recognizable.

3) Combined capital letters with the so-called "prosgegrammeni" (e.g.
U+1F88 to U+1F8F) have been designed as capitals with "ypogegrammeni",
just like in the charts of the Unicode Consortium. There is a major
issue here, but I had no choice anyway due to the matrix limitations.
Those who are familiar with those characters will know what to do; the
rest will not care.

4) For the Coptic letters I have used the charts of the Unicode
Consortium as a model.

---------------------------------------------------------------------------
From: "Constantine Stathopoulos" <cstath@irismedia.gr>
Date: Mon, 16 Aug 1999 19:41:26 +0300
Subject: Greek phi mixup

Markus Kuhn wrote: 
> What troubles me a bit is that you have U+03C6
> and U+03d5 exchanged compared to how they are shown in both the ISO
> 10646 and Unicode standards. This might confuse some people, especially
> TeX users. How do other Unicode fonts (e.g., Microsoft) handle this?

ELOT's opinion (mine and other Greeks', too) has always been that
characters U+03D0 to U+03D6 and U+03F0 to U+03F3 are just glyph
variations and should NOT have been included in the standard. As it
is, however, one should put the basic (most used) glyph in U+03C6 (or
U+03B2, U+03B8, etc.) and the alternative (less used) glyph in U+03D5
(or U+03D0, U+03D1, etc.). In the case of PHI the open glyph is used
in 95% of fonts, so my choice reflects the way the Greeks print their
texts. Monotype's WGL4 fonts (MS Windows Times, Arial, Courier) also
use the open PHI glyph, since they have been designed after old Greek
Monotype fonts. On the other hand, Monotype's Arial MS Unicode
(distributed with Office 2000) treats PHI the other way round;
however, Arial MS Unicode is a test Unicode font, not a real practice
font and has been designed by copying the images in the Unicode
charts. Its designers were probably not well familiar with the
Greek script.

[...]

I sent a paper to Asmus Freytag some time ago on his request. It is
possible that the images/glyphs will be switched in Unicode 3.0.
Anyway, feel free to bring the matter to the Unicode list, if you
wish.

For these and other issues, I would highly recommend Dr. Haralambous'
"From Unicode to Typography, a Case Study: the Greek Script,
Proceedings of the 11th Unicode Conference, Boston, 1999" available at
<http://genepi.louis-jean.com/omega/boston99.pdf>. (Caveat: the file
is 4 MB big!)

Dr. Haralambous is a Doctor of Mathematics and a TeX expert (co-author
of Omega). A significant part of his paper is dedicated to Greek in
Mathematics.

---------------------------------------------------------------------------
Date: Wed, 18 Aug 1999 20:15:26 -0700
From: Asmus Freytag <asmusf@ix.netcom.com>
Subject: Re: Greek phi mixup

This has been reported before, and we have independently verified that
other implementations from different and competing major vendors also 'fix'
this one quietly. Therefore these glyphs will be swapped Unicode 3.0 and
the next printing of ISO 10646.

This is an editorial correction of misplaced glyphs, not a change in
character assignment. The fact that so many organizations and individuals
independently concluded that what was there must be wrong and fixed it the
same way underscores that the nature of the charaters themselves was
sufficiently obvious from context and character name.
---------------------------------------------------------------------------
From: Birger Langkjer <birger.langkjer@image.dk>
Date: Tue, 21 Sep 1999 16:17:11 +0200

After experiencing some critism of the Unicode charts, I decided to
redesign the armenian glyphs for helvR12 based on a chart I found on
http://moon.yerphi.am/~hovik/Armenian/ArmSCII-7.gif

Unless someone finds a better chart or finds some faults with it,
these glyphs should be canonical, and the other fonts should be made
to reflect them IMHO.
---------------------------------------------------------------------------
From: Theppitak Karoonboonyanan <thep@links.nectec.or.th>

The Unicode 2.0 book is not quite good a reference for Thai glyphs. I
found the ones in ISO/IEC 10646-1:1993 (first edition) much more
perfect.
---------------------------------------------------------------------------
From: Serge Winitzki <S.Winitzki@damtp.cam.ac.uk>
Date: Fri, 15 Oct 1999 21:20:05 +0100 (BST)
Subject: Cyrillic issues

Cyrillic letters occupy 0400 to 04FF.

About the "historic" Cyrillic characters: The following characters are
very, very historic and obsolete (i.e. basically only used in research
on pre-1700 texts): 0460, 0461, 0464--046F, 0476--0486.

The characters 0462, 0463, 0470--0475 were still in use in 1900 and
some books used 0462, 0463, 0472, 0473 actually as late as 1940
(outside of the USSR). I would consider the latter four characters as
still marginally useful (e.g. for quotations) although the
contemporary Russian does not use them.

About shapes of individual letters:

U+0431 Cyrillic small be: make sure it's either a small version of
U+0411 Cyrillic capital Be, or an alternative shape that must be
distinct from the digit 6.

U+0414, U+0434 Cyrillic De: although it's of Greek "delta" origin, it
does not need to be triangular at all; in fact it is not triangular in
most contemporary fonts. It should look more like U+041B, U+043B
Cyrillic EL on top of a clockwise rotated '[' character.

U+0417, U+0437 Cyrillic Ze: make sure it's distinct from Cyrillic E
and from digit 3 (although it should rather resemble the latter).

U+043A Cyrillic small Ka: must have "x height" (unlike Latin "k") but
otherwise is very similar.

U+041B, U+043B Cyrillic EL: make sure it's distinct from U+041F,
U+043F Cyrillic Pe, either by the ascender at left, or by a slightly
smoother shape of its top.

U+041F, U+043F Cyrillic Pe: both capital and lowercase versions must
be of the same shape as U+03A0 Greek Capital Pi.

U+0444 Cyrillic small Ef: the lowercase Ef must have a stem that
extends below the line, and above to "cap height".

U+0426, U+0429, U+0446, U+0449 Cyrillic Tse and Shcha: the descender
should, if possible, be attached to the right of the letter. If not
possible (small fonts, letter Shcha), it's ok to have it below the
rightmost vertical line.

U+042A, U+044A Cyrillic hard sign: if possible, make the top line
larger, since it's the only distinction from U+042C, U+044C Cyrillic
soft sign.

U+042B, U+044B Cyrillic Yeru: if font size is small, it is permissible
for the two disjoint pieces to touch.

U+0409, U+0459 Cyrillic LJE: since it's a combination of Cyrillic EL
and Cyrillic soft sign, its left portion should not look like Cyrillic
Pe but rather like Cyrillic El (when possible).

U+0462, U+0463 Cyrillic Yat: the lower portion of the letter should be
exactly like Cyrillic soft sign, the height of the dash should be the
same as the "x height", and the stem should extend to "cap height"
above it.

U+0472 U+0473 Cyrillic small fita: it should have "x height" (unlike
its parent, the lowercase Greek "theta", which is of "cap height"),
essentially it is an "o" with a dash inside. It is not really
necessary to have a broken dash line there either.
---------------------------------------------------------------------------
From: Markus Kuhn <Markus.Kuhn@cl.cam.ac.uk>
Date: 2000-12-07
Subject: Terminal characters

Background information on the new terminal emulator characters
in Unicode 3.2 can be found in

  ftp://kermit.columbia.edu/kermit/ucsterminal/ucsterminal.txt
  ftp://kermit.columbia.edu/kermit/ucsterminal/terminal-exhibits.pdf
  http://www.columbia.edu/kermit/standards.html
---------------------------------------------------------------------------
From: jg@pa.dec.com (Jim Gettys)
Date: Tue, 29 May 2001 10:05:24 -0700
Subject: Re: history of -misc-fixed-* fonts

> Do you have any recollection where 6x13 and the other -misc-fixed-*
> fonts came from originally? Who made them or who might know who did?

I don't honestly remember, for sure...  They may have come off of the VS100's
that X first run on.  They may have been freely available fonts from that
era.

I'd be surprised if Bob's memory was any better than mine on the topic.

--
Jim Gettys
Technology and Corporate Development
Compaq Computer Corporation
jg@pa.dec.com
---------------------------------------------------------------------------
Subject: Re: history of -misc-fixed-* fonts
From: Bob Scheifler - SMI Software Development <rws@east.sun.com>
Date: Fri, 1 Jun 2001 15:26:25 -0400

> Do you have any recollection where 6x13 and the other -misc-fixed-*
> fonts came from originally? Who made them or who might know who did?

My memory of who did what fonts is gone, but here's what
Stephen Gildea has to say:

I think I did once know who wrote the fonts, but I've forgotten now.

The classics 6x10, 8x13 and 9x15 may have come from DEC.
They have DEC VT100 drawing characters in the 1-31 range.

I remember 6x13 was added in R4.

I myself wrote 5x7 and the ASCII portions of 7x13 and 7x13B.

Thomas Bagli of Germany did the Latin 1 extension for 6x13, 7x13,
8x13, 9x15, and their bold counterparts.
I wrote the Latin 1 for nil2, 6x10, and 10x20.

NCD contributed the ASCII part of 10x20.  I think Jim Fulton wrote it.

Don Knuth (!) contributed tweaks to 9x15B.

- Bob
---------------------------------------------------------------------------