Loading ...
Sorry, an error occurred while loading the content.

Re: GNU Unicode Font, Arabic Glyphs (fwd)

Expand Messages
  • Behdad Esfahbod
    ... My last mail to gnu-unifont@yahoogroups.com failed, unfortunatly I didn t tried to findout where the problem is. ... You were right, then I tried to have a
    Message 1 of 6 , Jul 10, 2001
    View Source
    • 0 Attachment
      On Tue, 10 Jul 2001, David Starner wrote:

      > Thanks! In the future, you should probably send stuff like this to
      > gnu-unifont@yahoogroups.com, since that's where all the gnu-unifont people
      > are;

      My last mail to gnu-unifont@yahoogroups.com failed, unfortunatly I
      didn't tried to findout where the problem is.

      > I doubt you've had time to really look at the new Unifont yet. Mikka-Markus
      > Alhonen provided quite a few new Arabic glyphs for the release. Since, as I
      > said, Arabic is out of knowledge range, I'd like him to look them over.
      > Mikka-Markus?

      You were right, then I tried to have a look at them, there is also a
      problem in the distribution, I guess you worked with it with cygwin
      ..., when I failed to convert the hex file to bdf, found that the
      regular expressions of the perl script does not match, and found that
      the lines of unifont.hex are terminated by CR/LF, instead of regular
      LF, after that, had a look at glyphs:

      1. U+262B IRI SIGN is very very strange, also I will send you
      a double-width one.
      2. Arabic glyphs are not still goodlooking to me, but there is
      one point that you have some ligatures added too, if you
      are not going to use my glyphs, I will try to at least
      improve yours.

      --
      Behdad
      19 Tir 1380, 2001 Jul 10

      [Finger for Geek Code]
    • Miikka-Markus Alhonen
      Hello! Sorry that my answer took so long, but I ve been rather busy with my day-time job lately :(. First I would comment on some miscellaneous stuff on the
      Message 2 of 6 , Jul 30, 2001
      View Source
      • 0 Attachment
        Hello!

        Sorry that my answer took so long, but I've been rather busy with my day-time
        job lately :(.

        First I would comment on some miscellaneous stuff on the new release of Unifont.

        Following the pattern of all the other Latin letters having dots above (U+010A,
        U+010B, U+0116, U+0117, etc.), the new additions in the range U+0226-U+0231
        should also represent the dot as a short vertical line of two pixels (much more
        noticeable than just a single pixel) in column 4 (one dot) or columns 3 and 6
        (diaeresis).

        0226:1010000018242442427E424242420000
        0227:0000101000003C42023E4242463A0000
        022B:003C002424003C4242424242423C0000
        022E:101000003C42424242424242423C0000
        022F:0000101000003C4242424242423C0000
        0230:3C0010003C42424242424242423C0000
        0231:003C001010003C4242424242423C0000

        The representation of diaeresis applies also to newly added Cyrillic letters:
        04EC:24240000384402023E02020244380000
        04ED:0000242400003844023E020244380000

        The macron in other characters (e.g. U+0112, U+0113) is placed two rows above
        the base character. This should be the case for the following characters, too:

        0232:003E0000414122221408080808080000
        0233:0000003C00004242424242261A02023C

        According to Unicode charts there should be two dots instead of a vertical line
        around the tilde in U+034B:

        034B:0800324C001000240042002400000000

        The present appearance of the question mark as a part of U+2048 and U+2049 is
        quite different from that of U+003F. I think they would be better like this:

        2048:00000000394545050911110011110000
        2049:000000004E5151414244440044440000


        Then something else to be considered:

        The double diacritics U+0360-U+0362 are meant to extend over or below two
        single-width characters, so wouldn't it be reasonable to have their glyphs
        double-width also?

        Since most of the Hebrew vowel points, cantillation marks etc. U+0591-U+05C4
        take a lot of space below the character, all the Hebrew letters in
        U+05D0-U+05F2 and U+FB1F-U+FB4F should be lifted up by two rows, so that
        programmes like Yudit could intelligently place the diacritics simply by
        overstriking and the result would be readable. Any opinions? The circles in the
        combining characters should be removed, of course.


        Finally, to the Arabic glyphs.

        On 10-Jul-01 David Starner wrote:

        > Behdad Esfahbod <behdad@...> writes:
        >> 0600-06FF: whole Arabic block

        The Koranic annotation signs U+06D6-U+06ED can well be accepted as they are,
        since we have no alternative versions for them.

        In general, the design of Behdad's Arabic blocks is much nicer than the original
        Roman Czyborra's ones, which I modified. I think most of them can be accepted as
        the new version, but there are a few things I would like to comment.

        First of all, it was a great idea to move the baseline of characters up by two
        rows, since now there are a lot less horrible-looking characters, cf. the old
        and new versions of U+FBE5. When I drew my version, I didn't want to change the
        whole design of the Arabic block, just add new letters, so this didn't even
        cross my mind.

        Also, I didn't realise before, that unlike in most other Unicode scripts, in
        Arabic you don't have to have an empty first or last column except for Alif,
        Dal, Ra, Waw etc., because of the joining properties of the characters. By
        drawing many letters as wide as possible, you can get much more naturality to
        their shapes.

        There are, however, a few minor things still bothering me in Behdad's version:

        The semicolon U+061B, decimal point U+066B and numbers U+0661-U+0669 &
        U+06F1-U+06F9 are all a bit thick, as compared to the actual letters.
        Independently they look very nice but in the middle of a text they all look
        like bolded characters.

        The shape of the extended number zero U+06F0 also amazes me. Do you really use
        that kind of an empty circle in Persian? In Urdu, at least, it's just the same
        as U+0660.

        Do you use something like the Nasta`liq style in Persian? I ask this because
        the base of Sin and its derivatives looks almost like a straight line. In the
        Naskh style, which is quite common in printing Arabic, the hooks of Sin should
        be clearly visible; at least 2 pixels in this case. Also the base of the letter
        should remain on the level of the baseline in the final and isolated forms,
        too. So, my suggestion for Sin would be something like this (the final and the
        isolated forms are a bit difficult if you want a nice-looking result):

        FEB1:00000000000115151A88848478000000
        FEB2:00000000000015151A88848478000000
        FEB3:0000000000011515EA00000000000000
        FEB4:0000000000002A2AD500000000000000

        The medial form of Ghayn U+FED0 and `Ayn U+FECC look too much like the medial
        form of Fa U+FED4 and its dotless version, I think. Could the former ones be
        squeezed a little bit to a narrower space? I don't know about Persian, but at
        least in most Arabic fonts the part above the baseline is just a big black
        spot, much like the one in the original version of the font.

        The different contextual forms of Kaf and letters derived from it U+FB8E-U+FB9D
        seem to have their topmost line in many different angles. I think they should
        all be the same, in order to not to give a confusing impression.

        The final form of Mim U+FEE2 looks quite weird, as the loop is below the
        baseline. Again, in Naskh the final form is often almost identical to the
        isolated form, which is usually like the one in your version or the one in the
        original version of the block.

        The initial form of Ha U+FEEB looks a bit strange, because usually in Arabic the
        slanted stroke in the beginning is much more upright, more like the one in the
        original version of the font. Also the medial form U+FEEC looks strange when
        you have a V-like letter and a loop in the end of it. I have seen only versions
        looking like V without a loop, and those resembling to 8. In the Unicode charts
        the latter one is used, so maybe it could be the version in this font, too?
        Though, the non-contextual glyph for the letter U+0647 in the charts looks odd
        to a person knowing only Arabic (out of the languages using this letter);
        the normal glyph for me would be the isolated form of the letter U+FEE9. I
        think they have chosen the glyph just so that it wouldn't get mixed with ARABIC
        LETTER AE U+06D5. I don't think this is necessary, because there are in any
        case glyphs which will get mixed by their appearance, such as Alif maqsura
        U+0649 and dotless Ya U+06CC, and most likely the languages using the letter Ae
        don't use this kind of Ha.

        According to the Unicode charts and to my knowledge of Arabic, too, the sign of
        nunation Fathatan U+064B shouldn't be on the final Alif but on the preceding
        letter of it. I know that some languages which take Arabic borrowings, write it
        on the Alif, but shouldn't we follow the Arabic practice, because it's the
        origin of this script?

        And finally the clear mistakes in the glyphs:
        -The baseline of U+FB51 is one row too low
        -The base of U+FC5B should be Dhal U+0630, not Dal U+062F
        -U+FDF2 should have the Alif in the beginning, too, according to the Unicode
        charts. So it's impossible to fit that in to a single width cell.

        >> FB50-FDFF: all Arabic Presentation Forms B possible glyphs, ie.
        >> those that fit in a single width space

        I didn't draw too much of ligatures in this block, so I might as well start
        from the beginning, if the general design is decided to be stable now.

        >> FE50-FEFF: whole Arabid Presentation Froms A

        Actually the block begins at U+FE70, not U+FE50, so the characters in
        U+FE50-U+FE6F, which you, Behdad, redrew, are not supposed to be Arabic, but
        small form variants of CJK punctuation, i.e. they should indeed be double-width.

        >> 262B: IRI Sign

        I have no idea about this Iran symbol but I suppose Behdad has, as he's Iranian
        himself, right? And this one does resemble more to the one in the Unicode charts
        than the previous one did. BTW, there are 4 extra zeroes in the end of this
        glyph's hex code, so it should be apparently:

        262B:00000000542854AAAAAAAA5438000000

        > I doubt you've had time to really look at the new Unifont yet. Mikka-Markus
        > Alhonen provided quite a few new Arabic glyphs for the release. Since, as I
        > said, Arabic is out of knowledge range, I'd like him to look them over.
        > Mikka-Markus?

        Actually there are two i's in the first part of my name, i.e. Miikka, not
        Mikka. If you find it too long a name to write "Miikka-Markus", you can just
        leave the "Markus" part out.

        Best regards!

        ----------------------------------
        E-Mail: Miikka-Markus Alhonen <Miikka-Markus.Alhonen@...>
        Date: 31-Jul-01
        Time: 02:03:40

        This message was sent by XFMail
        ----------------------------------
      • Behdad Esfahbod
        Hi Miikka-Markus, I replying to a pretty old mail. I ve found some time to work on bitmap fonts again, with a new set of bold glyphs. Would you please send
        Message 3 of 6 , Apr 11 6:09 PM
        View Source
        • 0 Attachment
          Hi Miikka-Markus,

          I replying to a pretty old mail. I've found some time to work on
          bitmap fonts again, with a new set of bold glyphs. Would you
          please send me your latest version of Arabic glyphs (any font,
          any size?).

          Yours,
          Behdad

          On Tue, 31 Jul 2001, Miikka-Markus Alhonen wrote:

          > Hello!
          >
          > Sorry that my answer took so long, but I've been rather busy with my day-time
          > job lately :(.
          >
          > First I would comment on some miscellaneous stuff on the new release of Unifont.
          >
          > Following the pattern of all the other Latin letters having dots above (U+010A,
          > U+010B, U+0116, U+0117, etc.), the new additions in the range U+0226-U+0231
          > should also represent the dot as a short vertical line of two pixels (much more
          > noticeable than just a single pixel) in column 4 (one dot) or columns 3 and 6
          > (diaeresis).
          >
          > 0226:1010000018242442427E424242420000
          > 0227:0000101000003C42023E4242463A0000
          > 022B:003C002424003C4242424242423C0000
          > 022E:101000003C42424242424242423C0000
          > 022F:0000101000003C4242424242423C0000
          > 0230:3C0010003C42424242424242423C0000
          > 0231:003C001010003C4242424242423C0000
          >
          > The representation of diaeresis applies also to newly added Cyrillic letters:
          > 04EC:24240000384402023E02020244380000
          > 04ED:0000242400003844023E020244380000
          >
          > The macron in other characters (e.g. U+0112, U+0113) is placed two rows above
          > the base character. This should be the case for the following characters, too:
          >
          > 0232:003E0000414122221408080808080000
          > 0233:0000003C00004242424242261A02023C
          >
          > According to Unicode charts there should be two dots instead of a vertical line
          > around the tilde in U+034B:
          >
          > 034B:0800324C001000240042002400000000
          >
          > The present appearance of the question mark as a part of U+2048 and U+2049 is
          > quite different from that of U+003F. I think they would be better like this:
          >
          > 2048:00000000394545050911110011110000
          > 2049:000000004E5151414244440044440000
          >
          >
          > Then something else to be considered:
          >
          > The double diacritics U+0360-U+0362 are meant to extend over or below two
          > single-width characters, so wouldn't it be reasonable to have their glyphs
          > double-width also?
          >
          > Since most of the Hebrew vowel points, cantillation marks etc. U+0591-U+05C4
          > take a lot of space below the character, all the Hebrew letters in
          > U+05D0-U+05F2 and U+FB1F-U+FB4F should be lifted up by two rows, so that
          > programmes like Yudit could intelligently place the diacritics simply by
          > overstriking and the result would be readable. Any opinions? The circles in the
          > combining characters should be removed, of course.
          >
          >
          > Finally, to the Arabic glyphs.
          >
          > On 10-Jul-01 David Starner wrote:
          >
          > > Behdad Esfahbod <behdad@...> writes:
          > >> 0600-06FF: whole Arabic block
          >
          > The Koranic annotation signs U+06D6-U+06ED can well be accepted as they are,
          > since we have no alternative versions for them.
          >
          > In general, the design of Behdad's Arabic blocks is much nicer than the original
          > Roman Czyborra's ones, which I modified. I think most of them can be accepted as
          > the new version, but there are a few things I would like to comment.
          >
          > First of all, it was a great idea to move the baseline of characters up by two
          > rows, since now there are a lot less horrible-looking characters, cf. the old
          > and new versions of U+FBE5. When I drew my version, I didn't want to change the
          > whole design of the Arabic block, just add new letters, so this didn't even
          > cross my mind.
          >
          > Also, I didn't realise before, that unlike in most other Unicode scripts, in
          > Arabic you don't have to have an empty first or last column except for Alif,
          > Dal, Ra, Waw etc., because of the joining properties of the characters. By
          > drawing many letters as wide as possible, you can get much more naturality to
          > their shapes.
          >
          > There are, however, a few minor things still bothering me in Behdad's version:
          >
          > The semicolon U+061B, decimal point U+066B and numbers U+0661-U+0669 &
          > U+06F1-U+06F9 are all a bit thick, as compared to the actual letters.
          > Independently they look very nice but in the middle of a text they all look
          > like bolded characters.
          >
          > The shape of the extended number zero U+06F0 also amazes me. Do you really use
          > that kind of an empty circle in Persian? In Urdu, at least, it's just the same
          > as U+0660.
          >
          > Do you use something like the Nasta`liq style in Persian? I ask this because
          > the base of Sin and its derivatives looks almost like a straight line. In the
          > Naskh style, which is quite common in printing Arabic, the hooks of Sin should
          > be clearly visible; at least 2 pixels in this case. Also the base of the letter
          > should remain on the level of the baseline in the final and isolated forms,
          > too. So, my suggestion for Sin would be something like this (the final and the
          > isolated forms are a bit difficult if you want a nice-looking result):
          >
          > FEB1:00000000000115151A88848478000000
          > FEB2:00000000000015151A88848478000000
          > FEB3:0000000000011515EA00000000000000
          > FEB4:0000000000002A2AD500000000000000
          >
          > The medial form of Ghayn U+FED0 and `Ayn U+FECC look too much like the medial
          > form of Fa U+FED4 and its dotless version, I think. Could the former ones be
          > squeezed a little bit to a narrower space? I don't know about Persian, but at
          > least in most Arabic fonts the part above the baseline is just a big black
          > spot, much like the one in the original version of the font.
          >
          > The different contextual forms of Kaf and letters derived from it U+FB8E-U+FB9D
          > seem to have their topmost line in many different angles. I think they should
          > all be the same, in order to not to give a confusing impression.
          >
          > The final form of Mim U+FEE2 looks quite weird, as the loop is below the
          > baseline. Again, in Naskh the final form is often almost identical to the
          > isolated form, which is usually like the one in your version or the one in the
          > original version of the block.
          >
          > The initial form of Ha U+FEEB looks a bit strange, because usually in Arabic the
          > slanted stroke in the beginning is much more upright, more like the one in the
          > original version of the font. Also the medial form U+FEEC looks strange when
          > you have a V-like letter and a loop in the end of it. I have seen only versions
          > looking like V without a loop, and those resembling to 8. In the Unicode charts
          > the latter one is used, so maybe it could be the version in this font, too?
          > Though, the non-contextual glyph for the letter U+0647 in the charts looks odd
          > to a person knowing only Arabic (out of the languages using this letter);
          > the normal glyph for me would be the isolated form of the letter U+FEE9. I
          > think they have chosen the glyph just so that it wouldn't get mixed with ARABIC
          > LETTER AE U+06D5. I don't think this is necessary, because there are in any
          > case glyphs which will get mixed by their appearance, such as Alif maqsura
          > U+0649 and dotless Ya U+06CC, and most likely the languages using the letter Ae
          > don't use this kind of Ha.
          >
          > According to the Unicode charts and to my knowledge of Arabic, too, the sign of
          > nunation Fathatan U+064B shouldn't be on the final Alif but on the preceding
          > letter of it. I know that some languages which take Arabic borrowings, write it
          > on the Alif, but shouldn't we follow the Arabic practice, because it's the
          > origin of this script?
          >
          > And finally the clear mistakes in the glyphs:
          > -The baseline of U+FB51 is one row too low
          > -The base of U+FC5B should be Dhal U+0630, not Dal U+062F
          > -U+FDF2 should have the Alif in the beginning, too, according to the Unicode
          > charts. So it's impossible to fit that in to a single width cell.
          >
          > >> FB50-FDFF: all Arabic Presentation Forms B possible glyphs, ie.
          > >> those that fit in a single width space
          >
          > I didn't draw too much of ligatures in this block, so I might as well start
          > from the beginning, if the general design is decided to be stable now.
          >
          > >> FE50-FEFF: whole Arabid Presentation Froms A
          >
          > Actually the block begins at U+FE70, not U+FE50, so the characters in
          > U+FE50-U+FE6F, which you, Behdad, redrew, are not supposed to be Arabic, but
          > small form variants of CJK punctuation, i.e. they should indeed be double-width.
          >
          > >> 262B: IRI Sign
          >
          > I have no idea about this Iran symbol but I suppose Behdad has, as he's Iranian
          > himself, right? And this one does resemble more to the one in the Unicode charts
          > than the previous one did. BTW, there are 4 extra zeroes in the end of this
          > glyph's hex code, so it should be apparently:
          >
          > 262B:00000000542854AAAAAAAA5438000000
          >
          > > I doubt you've had time to really look at the new Unifont yet. Mikka-Markus
          > > Alhonen provided quite a few new Arabic glyphs for the release. Since, as I
          > > said, Arabic is out of knowledge range, I'd like him to look them over.
          > > Mikka-Markus?
          >
          > Actually there are two i's in the first part of my name, i.e. Miikka, not
          > Mikka. If you find it too long a name to write "Miikka-Markus", you can just
          > leave the "Markus" part out.
          >
          > Best regards!
          >
          > ----------------------------------
          > E-Mail: Miikka-Markus Alhonen <Miikka-Markus.Alhonen@...>
          > Date: 31-Jul-01
          > Time: 02:03:40
          >
          > This message was sent by XFMail
          > ----------------------------------
          >

          --
          Behdad Esfahbod 23 Farvardin 1382, 2003 Apr 12
          http://behdad.org/ [Finger for Geek Code]

          In the corner of the dream was the man with the blue guitar
          It had no strings but the music touched the stars
        • alhonenm
          Salaam! Sorry for the delay in the response. ... As far as I recall correctly, the ones at Gáspár s site (unifont.hex combined with arabforms.hex) are the
          Message 4 of 6 , Apr 23 3:44 AM
          View Source
          • 0 Attachment
            Salaam!

            Sorry for the delay in the response.

            --- In gnu-unifont@yahoogroups.com, Behdad Esfahbod <behdad@b...>
            wrote:
            > I replying to a pretty old mail. I've found some time to work on
            > bitmap fonts again, with a new set of bold glyphs. Would you
            > please send me your latest version of Arabic glyphs (any font,
            > any size?).

            As far as I recall correctly, the ones at Gáspár's site (unifont.hex
            combined with arabforms.hex) are the latest ones. Please see
            http://www.yudit.org/download/fonts/bitmap/

            Best regards,
            Miikka
          Your message has been successfully submitted and would be delivered to recipients shortly.