Re: [HIG] Localized Guidelines

[Christian Rose <menthos gnome org>]

fre 2003-05-30 klockan 03.21 skrev Gregory Merchan:
> > > Images 3 and 4 show pathological cases. Most commands are and should be
> > > simple imperatives. Translations should follow the same implicit rule
> > > for verb selection that English does: use verbs of less than about 7
> > > letters.
> > 
> > For some languages, that's not even remotely possible. Many other
> > languages are also more verbose by their nature than English, so even
> > "short" imperatives tend to be longer than their English equivalent.
> > 
> > So even with a translation recommendation of "keep as short as
> > possible", the guidelines and design also need to allow for, and work
> > with, longer commands and commands with greatly varying lengths.
> 
> I miswrote. The translations should follow a rule _like_ "7 or less", but
> adjusted for a common verb length.

As we are likely to not have resources (volunteers) to localize the HIG
into all or even most languages that we support anytime soon, not to
speak of getting hard data on common verb length or other details for
all these languages, I'd rather see that we avoid any hard numbers that
will need to be localized in this area and instead write "as short as
reasonably possible" or some such.


> The change could specify a lower limit
> as well. The important thing is to keep lengths as similar as possible

Certainly, I'm just saying that it won't always be possible.

Perhaps the numbers are needed to give some level of guideline, but it
should be pointed out that it only applies to English, and while the
messages should preferrably follow the same design when translated into
other languages (preferrably short but not too short and if possible
with not too variant lenghts), the numbers will be different.


> In English we have the common odd case of "OK" and "Cancel" - a four
> letter difference.

That example gets even more interesting when translated into other
languages. In German, for example, you have "OK" and "Abbrechen" - a
seven letter difference. And German still isn't the most verbose
language.


> The stock items should already cover most cases; any not covered should
> be added. Once we have a list of items, then we should go through look
> for actual large variance and eliminate it where possible.

That may work. You can report such issues through Bugzilla, using the
l10n product.

There's a possibility though that such variance issues will in many
cases cannot be fixed, but it may still be worthwhile to report such
issues in case they should just be an oversight.


> If this proves intractable, altering the guidelines is not necessary.
> We have the option of localized guidelines. OPENSTEP supports this with
> string and interface files included in bundle localization directories.

As I said previously, we are unlikely to have localizations of the HIG
into all or most languages anytime soon, so we should probably not make
us depend on it with including lots of English-specific advise into the
document, without marking it as such, and relying on localizations to
alter those when appropriate. It would be better to directly mark
English-specific advise as such, and provide more general guidelines for
translators in the same document if needed.


> Localized guidelines may be necessary already. "OK" as used in English is
> suitable for OK/Cancel dialogs, information alerts, and error alerts; though
> the meaning is slightly different. I see the Spanish translation is "Aceptar"
> and I suspect that this does not always convey the correct meaning.
> Grepping for msgstr.*_OK in gtk+/po, I see all of these locales using "OK":
> de, el, fi, hu, id, it, ja, mn, ms, nl, nn, no, pl, pt, pt_BR, ro, ru, sk,
> and sv.
>
> It may even be the case that a word, such as "OK", should not be translated.
> "OK" is used by (at least some) native speakers of Spanish and French to
> convey the same meaning as the English word. Similarly, though not relevent
> to any user interface I've seen, the Italian word "ciao" is recognized as
> a salutation and valediction by native speakers of many languages. The
> Spanish "salud" also enjoys similar widespread use. I don't know what
> bearing this has on UI, but it's something to check.

You're on a very dangerous path here. While using examples from some
languages to illustrate that there is a localization problem, and that
something needs localization for some languages, is often correctly used
as a motivation for some design decisions, using the reverse motivation
is rarely possible. I.e. one can rarely ever claim that something
linguistic is the same in all languages and use that as a motivation for
not using localization, simply because noone has that necessary
knowledge.

In fact, such reverse logic has historically been the source for the
most naïve decisions and horrid design mistakes and flaws ever when it
comes to localization. We can do better.


Christian