reducing l18n bloat (was Re: Linking against libjpeg ?)


Subject: reducing l18n bloat (was Re: Linking against libjpeg ?)
From: Paul Rohr (paul@abisource.com)
Date: Wed Apr 25 2001 - 21:25:43 CDT


At 07:58 PM 4/21/01 +0200, Paolo Molaro wrote:
>On 04/19/01 Paul Rohr wrote:
>> >If I had a thing to say about it, I would say make it a
>> >loadable module (and consequently make loadable modules
>> >work properly) but don't force it to be present.
>> >
>> >AbiWord ... 3M and growing fast
>>
>> I'd still love to see stats on how much of that bloat is due to:
>>
>> - more file formats
>> - more localizations
>
>This accounts for half a meg of binary size. Again, gettext
>is the right solution here on unix (and probably in other
>systems, too).

Thanks for the estimate. What's the breakdown on that half a meg between
file formats and localizations? If it's mostly file formats, then we may
want to work harder on externalizing impexp modules, instead.

As for gettext, rather than replaying prior conversations on this topic, I'd
like to see if we can move forward. We're all in agreement that the current
setup -- where all UI strings except menus and toolbars are externalized --
is not ideal. There are too many discrete places for translators to update,
and it requires a recompile to test portions of the localization.

Admittedly, one incremental step could be to also externalize menu and
toolbar strings -- using the current XP mechanism *or* gettext. However,
neither change would address the fact that we still don't have a good way to
externalize the following menu and toolbar localizations:

  - layouts of specific items
  - associated icons

Thus, there are three *orthogonal* problems that could be addressed here:

1. How do we get *everything* out of the binary, not just the strings?

2. Would working copies of gettext ported to all our non-unix platforms
give us a technical advantage over our current solution? Is it faster or
more efficient at run-time than parsing and loading XML strings files? Are
the PO files more compact and easier to distribute? Do they help us avoid
platform-specific charset issues?

3. An oft-cited advantage of using PO files is that, on unix at least,
there are some translator-friendly tools which make maintaining PO-format
translations easier. What would it take to get *those* tools ported to make
things easier for translators on other platforms?

Paul



This archive was generated by hypermail 2b25 : Wed Apr 25 2001 - 21:18:10 CDT