[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Reformulation of criticisms about console-setup switch (was: Re: Review of console-setup wrt D-I [very long])



Quoting Frans Pop (elendil@planet.nl):
> On Tuesday 02 June 2009, Christian Perrier wrote:
> > One of the release goals of Debian Installer for squeeze is dropping
> > the use of console-data keyboard mappings, to replace them by
> > console-setup [1].
> 
> Here are (finally) the results of my tests. Sorry for the delay; I will 


During two long train trips, I carefully read Frans mail as well as
the followups made by others.

In the following, I tried to reformulate the main criticisms in a few
categories. The point is checking whether we share our understanding
of the problems...and throw out a few vague ideas about possible ways
to address these concerns.


Size impact
-----------

General concern.

Possible solutions:
- solve other issues...particularly the one about too many choices
  and the one about useless questions.

Special handling of translations
--------------------------------

Translations of keyboard layout, model, variant are handled in
/usr/share/c-s-mini/c-s.config. That brings a significant impact over
handling them in debconf as it doesn't allow dropping translations to
save memory as it's donne in low memory situations.

My understanding of the rationale for this: terms for these choices
come from xkeyboard-config and are translated there. So, the first
proposed implementation has been to take the strings and translations
from there and inject them in the config script.

Possible solutions:
- build the templates files in a similar way to localechooser and iso-codes, taking translations from xkeyboard-config .mo files
- ?

Offering keymaps for all architectures
--------------------------------------

In expert mode, keymap models for all arches are proposed.

Possble solutions:
- clean this out? :-)

Too many choices
----------------

The c-s udeb proposes everything proposed by xkeyboard-config. That
makes way too many possible combinations, therefore confusing choices
and a size impact (because of i18n).

Possible solutions:
- instead of a "everything possible in xkb-data", simplify the choices list by reducing to "one keymap, eventually two, for each supported language"

Wording problems for some choices
---------------------------------

"USA" vs "American English"...

We actually *already* had this debate about console-data and the clear conclusion was that using country names is very often wrong: keymaps are most often designed by the language they're designed for, first, then *eventually* differenciated by country names.

Possible solutions:
- change names..:-)...which actually means moving away from xkeyboard-config names and use our own naming scheme, eventually re-using names from console-data, that have proven being accepted for years.

Useless questions
-----------------

Some questions were not formerly asked and seem generally useless
*even in expert mode*.

Questioned templates: altgr, compose, encoding, charset, console font, font size, virtual consoles

Possible solutions:
- discuss about them and drop the useless questions (at least in D-I). Yes,
  that may mean not having the same codebase in D-I and outside D-I but
  we could also wonder whether those questions are relevant for standard
  installation of the package, anyway.


Preseeding broken
-----------------

That issue seems related to "not able to configure the keyboard twice".

Possible solutions:
- fix the "configure twice" problem
- really test preseeding


Allow skipping keyboard configuration
-------------------------------------

There's a regression here.

Possible solutions:
- add a medium priority question allowing to *not* configure the keyboard just
  like kbd-chooser is doing. One really needs to first check
  how that's done in kbd-chooser (which I can't as I'm anything but
  fluent in C)

Skip configuration for Serial and UML installs
-----------------------------------------------

Another regression.

Possible solutions:
- probably related to "Allow skipping...". Needs to use 
  debian-installer/serial-console and debian-installer/uml-console

Compose key does not work in D-I
--------------------------------

"already reported": find bug number

Attachment: signature.asc
Description: Digital signature


Reply to: