Do not ship translations without matching input support

[Martin Pitt]

We currently ship a lot of translations on the CDs, but some of them require special input support which is just available as l-support-* dependency at the moment. We should make sure that all necessary input support is present on the CD.

[Martin Pitt (pitti)]

For Chinese we need:
  libchewing3 libchewing3-data scim-chewing scim-modules-table scim-pinyin scim-tables-zh
which need 12 MB.

For Japanese:
  anthy libanthy0 scim-anthy
(4 MB)

and for a dozen other languages we need scim-tables-additional (400 kB). Of those languages, Bengali, Arabic, Hindi, and Russian are currently shipped on the live system.

This leaves us with three options that make some sense:
  - Ship Chinese input support and remove Portugese, Russian, Japanese, German, and French translations.
  - Drop Chinese translations from the CDs, add input support for Japanese and Bengali/Russian/Arabic/Hindi (scim-tables-additional).
  - Drop translations for all languages that require special input support and ship translations for languages that do not require special scim tables.

After speaking with Tollef, these changes are too disruptive for Herd-5, will do the change immediately afterwards.

[Martin Pitt (pitti)]

Matt, Tollef, do you have any preference which option we should take?

[Matt Zimmerman (mdz)]

On Tue, Feb 27, 2007 at 10:21:26AM -0000, Martin Pitt wrote:
> Matt, Tollef, do you have any preference which option we should take?

Are there any statistics we can draw on to help us decide, based on which
translations/input methods are being used in the field?

--
 - mdz

[Martin Pitt (pitti)]

I updated the langpacksize script to include the size of the necessary input support packages. Also, if you call it with 'MB' as first argument, it prints out sizes in MB instead of bytes.

The script is in langpack-o-matic's bzr, but I attach it here for convenience.

Please NB that the numbers are dependant on the priority order in some cases. E. g. 8 languages require scim-tables-additional, but since Bengali (bn) has the highest priority, the script adds the size for scim-tables-additional to Bengali only. If we would move the priority of e. g. Arabic higher (priority_langs array at top of the script), then the scim-tables-additional package size would be added only to Arabic instead. This makes sense because we add langpacks to the CD according to the same prioritization.

FYI, the first two handful of output lines, as of March 6 (this will change significantly when we'll actually get Feisty language packs):

$ ./langpacksize MB
en G: 5.26 K: 5.67 G+K: 8.09 GSum: 5.26 KSum: 5.67 G+KSum: 8.09
zh G: 16.57 K: 17.90 G+K: 20.81 GSum: 21.83 KSum: 23.58 G+KSum: 28.90
es G: 3.48 K: 4.80 G+K: 6.14 GSum: 25.31 KSum: 28.37 G+KSum: 35.04
bn G: 0.89 K: 0.89 G+K: 1.60 GSum: 26.20 KSum: 29.26 G+KSum: 36.64
hi G: 0.81 K: 1.32 G+K: 2.04 GSum: 27.01 KSum: 30.58 G+KSum: 38.69
ar G: 0.57 K: 1.12 G+K: 1.60 GSum: 27.58 KSum: 31.70 G+KSum: 40.29
xh G: 0.59 K: 0.09 G+K: 0.60 GSum: 28.18 KSum: 31.79 G+KSum: 40.89
pt G: 4.94 K: 7.72 G+K: 10.14 GSum: 33.11 KSum: 39.51 G+KSum: 51.04
ru G: 2.52 K: 3.83 G+K: 5.07 GSum: 35.64 KSum: 43.35 G+KSum: 56.10
ja G: 6.32 K: 7.59 G+K: 8.72 GSum: 41.95 KSum: 50.94 G+KSum: 64.82
de G: 3.28 K: 4.63 G+K: 5.97 GSum: 45.23 KSum: 55.56 G+KSum: 70.79
fr G: 3.76 K: 5.07 G+K: 6.44 GSum: 48.99 KSum: 60.63 G+KSum: 77.24

(These are all of our 'priority languages', i. e. world's most spoken languages).

[Martin Pitt (pitti)]

The question has been resolved now, langpacksize now takes those packages into account, and the current seeds do not ship any non-English translations for now. Tomorrow (when we'll have new live images) I'll add language packs as they fit, including necessary input support.