Re: Why people still want to encode precomposed letters

From: Hans Aberg (haberg@math.su.se)
Date: Sun Nov 23 2008 - 14:59:27 CST

  • Next message: philip chastney: "RE: Why people still want to encode precomposed letters"

    On 23 Nov 2008, at 20:17, Asmus Freytag wrote:

    >> It is in fact even more complicated, because even it is called
    >> LATIN CAPITAL LETTER A WITH RING ABOVE
    >> one can have, as a matter of rendering style, the ring dipping
    >> into the main character, even being partly obscured. So one may
    >> not get around recognizing certain combination and provide special
    >> designs for those.

    > Actually. I suspect that the reason this overlapping exists and is
    > considered acceptable has to do with generic factors, in this case,
    > the problem of placing a rather "fat" diacritic on top of a tall
    > character. To render a disconnected ring, you would need four
    > separate areas of alternating black and white, from the top of the
    > line to the top of the character (A, in this case). At small sizes,
    > or low resolution, it's hard to find that much space. Hence the use
    > of overlapping forms, which reduce the space requirement by half
    > (or rather, allow the ring to be twice the size it would otherwise
    > have to be).

    Looking into a high quality book from the time before computer
    typesetting, "Bonniers Lexikon", 1967, which is typeset using
    Caledonia from the Mergenthaler Linotype Company in 8 points, it does
    have in both regular and boldface the circle dipping into the "A",
    but not as much as the point upwards on the latter would stick into
    the ring. In addition, the ring seems to be drawn slightly to the
    right relative to this upper point - perhaps this is to compensate
    for the fatter right side of the letter.

    > With a bit of research, you could find other cases of such
    > adjustment and formulate layout principles for diacritics that
    > could help you do a better job in the absence of explicit designs.
    > These principles would result in a set of rules that's not as
    > simplistic as much that's been attempted, but I would predict that
    > when fully developed their complexity is not different from other
    > types of text-layout tasks.

    Yes, this might be viewed as a form of kerning. Kerning for a
    succession of characters can be done by recognizing certain character
    sequences, which ones may depend on the language. In modern times, it
    might be reasonable to find more general kerning formulas, perhaps
    language independent, but simpler to implement in a computer -
    recognizing pairs for a large number of characters might quickly
    become unfeasible. A model might be quite complicated, requiring say
    that the combining characters be drawn in layers, so that they are
    hidden by the main character of drawn towards very closely.

       Hans



    This archive was generated by hypermail 2.1.5 : Sun Nov 23 2008 - 15:02:02 CST