Subject: integrating bidi (was Re: version 1.0)
From: Paul Rohr (paul@abisource.com)
Date: Tue Feb 06 2001 - 11:32:07 CST
Presumably, we already have consensus on the following points:
- big patches are hard to review
- forcing Tomas to keep his patches in sync is burdensome
- bidi is definitely a desirable feature
- but it's also a performance and bug risk
Would it help to differentiate between:
- integrating a mature #ifdef ABI_BIDI patch, and
- turning that feature on for 1.0 binary users?
At this point, isn't the debate more about the first step than the latter?
If so, then I'd suggest that the two key questions are:
1. How well localized are the bidi changes?
2. Does anyone believe that Tomas' design is fundamentally unsound?
Adding a well-localized set of #ifdef ABI_BIDI changes to help speed up work
on the feature makes sense. Sprinkling random bidi dust all over the tree
is less appealing. (Think a dozen or two ifdefs, rather than an order of
magnitude more.)
This seems like a perfect time for the usual refactoring advocates to jump
in and help out. :-)
At 11:36 PM 2/6/01 +1100, Martin Sevior wrote:
>I REALLY like the idea of a 1.0 with bidi.
I love the idea of bidi binaries for the world to use. I'm less sure we'll
want to delay 1.0 for it.
Getting any significant new feature to the point that it Just Works takes
time. You can shorten the process considerably by careful design and
engineering up front, but there's always a substantial period of time needed
to "burn in" the feature and take any remaining rough edges off.
The 1.0 we already have in mind will knock people's socks off because it has:
- a useful, interesting feature set
- that Just Works, period,
- for so many languages,
- on so many platforms.
That alone is a very very big deal. When we hit that milestone, we'll have
something we can all take a huge amount of pride in, because literally
millions of people can and will be happy to use it.
Having another major release after that with tables that Just Work, or bidi
that Just Works, or outlining that Just Works will also be a big deal.
>There are lots of great
>Hewbrew hackers and a significant number of people who could use bidi
>support.
Absolutely. That's why we should focus first on getting working sources to
those hackers so they can help make them Just Work for all those users.
Paul
motto -- one step at a time
This archive was generated by hypermail 2b25 : Tue Feb 06 2001 - 11:24:33 CST