Re: [Evolution] i18n buglets

[Not Zed <notzed ximian com>]

On 28 Mar 2001 19:24:38 +0200, Ladislav Lhotka wrote:
> On 27 Mar 2001 16:42:08 -0500, Jeffrey Stedfast wrote:
>
>
> > > 2. With any Czech keyboard, the combination of the acute dead key and
> > > the letter 'o' should produce 'o with acute accent' (0xf3) but instead
> > > 'n with acute accent' (0xf1) comes out - this is probably a typo but I
> > > wasn't able to locate it. Other (even Gnome) apps don't have this
> > > problem.
> >
> > Maybe it's not using the right charset and/or font? I dunno...
>
> As everything else is correct, I bet it's just a copy-and-paste error.
>
> > > 3. In the header fields of the outgoing messages, the spaces between
> > > non-ascii words are lost. This
> > > is what happens to a Subject consisting of two Czech words:
> > >
> > > Subject: =?ISO-8859-1?Q?mal=FD?= =?iso-8859-2?Q?test=ED=E8ek?=
> > >
> > > ...
> >
> > The capitalization of the charset (iso-8859-2) doesn't matter, I'll look
>
> I know, but why not be consistent if it costs nothing? Actually, often
> we have to deal with such cryptic subjects directly (e.g.,
> pilot-mailsync doesn't decode it yet) and so it is also important that
> these quoted-printable subjects be short and readable as much as
> possible. So I would prefer the following encoded form of the above
> subject:
>
> Subject: =?ISO-8859-2?Q?mal=FD_test=ED=E8ek?=
>
> Pine and even Outlook do it exactly this way. Why use two charsets if it
> can be covered by one?


Hmm, this is actually rather hard to do with the way the code works.

What its trying to do for example is try and find words that are
us-ascii so it can just put them in without encoding when it can (i.e.
to aid readability), and you've just hit an edge case :(  (which i'm
sure comes up in practice often with these character sets).

That is possibly why it's dropping the space, since it only expects to
do this when one of the words aren't encoded (?).

Jeff, make sure you add this as a test case to the test code if you are
going to look at it, which badly needs updating again anyway.  Or I can
look at it if you want.  It might need a different/expanded approach,
i.e. if two words are disparate encodings, then try and merge subsequent
words and rescan them for the best charset to use, etc.  And that code
is already a shit.

 !Z