pruning the fonts list

Warren Togami wtogami at redhat.com
Fri Apr 18 22:35:35 UTC 2008


Nicolas Mailhot wrote:
> I'm afraid most wins are in 3. 5. 6. & 7., and they depend on l10n
> groups telling us their wishes, which unfortunately has not happened a
> lot so far. http://fedoraproject.org/wiki/SIGs/Fonts/Triaging/L10N
> remains terribly incomplete.
> 

Even if you had full knowledge of fonts safe to remove from the LiveCD
(because they are good enough for web browsing), this introduces an
additional problem of incomplete language coverage in Live installs.
This is not a problem if the LiveCD is used for one-off demos, but it is
an issue if you are using it to install your desktop.

This is a new source of user confusion, bogus support questions and
bogus bugs.  It isn't exactly obvious to the user that they need to use
"yum groupinstall foo-language" in order to obtain full support for
their native language.  There is also no way to do group installs from
PackageKit GUI at the moment.

This was not a previous behavior, this expectation isn't documented
anywhere, and this potentially invalidates F9 cycle testing because
testing did not happen with incomplete font package installs.  You
especially want to avoid culturally insensitive (and difficult for
outsiders to test) cases where the wrong font is giving glyphs for a
language it was not intended, as can happen between Chinese, Japanese
and Korean.

This is a big change to make this late in the F9 cycle.  Can we 
adequately test the impact of this to all supported languages?

Key question to ask ourselves: Is trying support all languages from a
single Live image really sustainable?  Even after removing all fonts
unnecessary for web-browsing, we cannot indefinitely continue to add an
arbitrary number of languages to a fixed size CD image.

We may be forced into doing region/language specific LiveCD's.

Let us get an expert opinion from RH's eng-i18n team.

Warren Togami
wtogami at redhat.com




More information about the fedora-devel-list mailing list