Re: random differences (was Re: selections and combining characters)

From: Paul Rohr (
Date: Thu Apr 25 2002 - 00:56:47 EDT

  • Next message: Martin Sevior: "Re: Pango?"


    Thanks for the summary. I'm not the guy you have to convince, though.

    So far, the bulk of our XP i18n and script-handling support has been coded
    by Tomas Frydrych. He's a practical guy. Sell *him*. :-) Given what he's
    managed to accomplish so far, and his continued willingness to keep moving
    forward, his take on these issues matters a lot more than mine.

    As I'm sure you realize, there's a subtle undercurrent in the argument
    you're making for our use of Pango. In short, Pango -- especially the
    lower-level stuff -- might indeed be ideal for our purposes, *if* it didn't
    have "too many" of the following issues:

      - coverage of less-common shaping engines,
      - portability,
      - printing, and (perhaps)
      - performance [1].

    As with adopting any library, the tradeoff we face is whether the benefit to
    us is worth the pain of integrating *and improving* somebody else's code
    now. [2]

    After all, one viable alternative (for us) is to simply wait until enough
    South Asian and Middle Eastern developers show up to do the necessary work
    in *either* of our code bases. [3] Indeed, that kind of recruiting strategy
    has worked for us quite well in the past. ;-)

    At 10:29 PM 4/24/02 -0400, Havoc Pennington wrote:
    >I also think you guys could make important enhancements to Pango.
    > [ ... ]
    >And then there's a feature it doesn't have yet, namely printing, that
    >you would want to use when it exists. Maybe you could write it. ;-)
    > [ ... ]
    >I'm quite sure there are small details of Pango that don't quite do
    >what you need, but those can be fixed.

    In a nutshell, it almost sounds like Pango needs us more than we need Pango
    -- for the moment, at least. Given the relative resources being invested in
    GNOME vs. AbiWord, that sounds backwards. ;-)

    >I don't know the right answer, I just find it hard to believe that
    >creating a whole new system for something this
    >fundamental/infrastructural is the right answer.

    Yep. I know where you're coming from. If there's a practical way for Owen
    and Tomas to easily leverage each other's work here, that'd be a Good Thing.

    bottom line
    What both of our communities face is a recruiting problem -- not enough
    developers have shown up who are ready, willing, and able to do the
    necessary work in this area. In the mean time, it's up to lone heroes like
    Owen and Tomas to do whatever makes the most sense to them.

    Compared to that, the rest of us are just lobbyists, right? :-)


    [1] Given some experiments we did with FreeType back in 98/99, I'm even a
    bit leery about performance issues just at that layer. Still, it's been a
    while, and Tomas has earned my trust, so if he says he can make it work,
    I'll believe him until proven otherwise.

    [2] If the benefit is big enough, we've historically shown an astonishing
    tolerance for pain. For example, I'm one of the long chain of designated
    suckers who collectively coerced a balky ispell into running on non-Unix
    platforms. As anyone who's touched that code knows, none of the following
    tasks were enjoyable:

      - upgrade it to ANSI C,
      - work around file format variations,
      - plug lots of nasty leaks,
      - add charset conversions,
      - make it reentrant, and
      - other tasks blissfully forgotten.

    If it weren't for all those tasty ispell dictionaries out there, trust me, we
    *never* would have done so.

    [3] I'm quite confident that once good code exists to do the job we both
    need, we'll all find ways to take advantage of it, regardless of who first
    wrote it where.

    This archive was generated by hypermail 2.1.4 : Thu Apr 25 2002 - 00:57:10 EDT