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 different from 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 . It is not supposed to look like one. It looks more like a ligature of a (U+017f) followed by a (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 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 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 . 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

and , 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 and , so if you have enough pixels, try making it look like that. U+0251, U+0252 (script a, turned script a) If the default in your font has is a , try to make this an exaggerated . If is not the same as , 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 , which might have a loop below. U+0264 (rams horn) This needs to be graphically distinct from , and . 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://partners.adobe.com/asn/developer/typeforum/unicodegn.html --------------------------------------------------------------------------- From: Birger Langkjer 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 (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" 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 . (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 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 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 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 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 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 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 ---------------------------------------------------------------------------