Re: IMPORTANT: proposed removal of non-bidi code

From: Scott Rushfeldt (sirushfe@unity.ncsu.edu)
Date: Wed May 08 2002 - 10:58:19 EDT

  • Next message: Scott Rushfeldt: "Re: IMPORTANT: proposed removal of non-bidi code"

    Paul Rohr wrote:
    > At 06:26 PM 5/7/02 +0200, Christian Biesinger wrote:
    > >Paul Rohr wrote:
    > >> 1. native, non-bidi ... the well-tested code that everyone uses now
    > >> 2. bidi ... some testing, not enough use
    > >> 3. Pango-based ... to be written and/or ported as needed
    > >[...]
    > >> I also am willing to believe that we'll get to the point where #3 is
    good
    > >> enough that we should *also* replace #1. However, I'm stunned to hear
    > that
    > >> we're already at this point.
    > >
    > >As I understand it, #2 will replace #1; not #3.
    >
    > Why do it then? The whole point of doing #3 is to replace #2, right? For
    > many many languages -- indeed for most existing computer users AFAIK -- #1
    > Just Works already and *neither* #2 nor #3 is needed. [**]

    If I understand things right #3 will not be replacing #2. All pango will be
    used for,
    from what I understand, is to render the correct gyphs on screen that
    correspond
    to the unicoded characters entered from the keyboard. While this might
    replace
    some of the Hebrew and other languages rendering code in the Bidi
    distribution
    it will not replace the Bidi layout and selection. Thus, if I am right,
    the reasoning
    for removing the non-bidi code is it does nothing more then the bidi code
    and is
    only slightly smaller making the keeping both versions more of a hassle than
    is
    worth. Also, with all the new features and refactoring of code to be done
    it would
    be a major hassle to have to put all the changes and debug them in both the
    bidi
    and non-bidi builds.

    > I understand that simultaneously maintaining all three variants is a
    > headache, but this sounds like we'd be dropping the wrong one now. If
    we're
    > going to have only two variants of the code for a while, wouldn't it make
    > more sense to drop #2 and focus our efforts on #1 (needs no work) and #3
    > (needs significant work)?
    >
    > By switching to #2, don't we expose ourselves to bugs in BiDi-related
    > portions of the codebase that *we plan to abandon* (in favor of Pango)?

    Yes, there might be more bugs, but from what I understand the bidi code is
    pretty
    stable now. As well, since we hope to some day say that Abi supports all
    the worlds
    languages it is necessary that the main build is bidi, otherwise Abi will
    not support
    all the worlds languages only Abi's bidi build will.

    >
    > I still must be missing something here.
    >
    > Paul
    >
    > [**] To be clear. I want #3 to move forward. Supporting more languages
    is
    > a very Good Thing. Leveraging other people's expert work to do so is also
    a
    > Good Thing. I'm just completely failing to see why we should abandon #1
    > before #3 is further along.
    >

    Hope what I've said is correct and helps answer your questions on why the
    non-bidi
    code is up for removal.

    Scott Rushfeldt



    This archive was generated by hypermail 2.1.4 : Wed May 08 2002 - 10:51:55 EDT