Committed, thanks.



The translated comments are rather for the end users than the administrators.

No it isn't, I know now :)



Anyway, now I read your coding_standards document (better lately, than 
never) and I found, that I should use include_once, so I fix this in 
this patch.



A bigger fix is utf8'ized comments, since it is very annoying if you 
can't pipe the generated output to a 'recode UTF-8..ISO-8859-2' as a 
system administrator because of ISO-8859-2 comments (which are invalid 
UTF-8 characters).



Anyway, I think, that it is not a good idea to have localized comments 
in a generated config file.  System administrators know english well 
enough and if someone is doing some shell magic based on english 
comments it can be a bg surprise that he has got non-english users 
with localized chinese autocommented sieve data... :)



So _() should be removed from around comment strings which will go to 
sieve scripts, I think.



Thanks,

Gergely

Isn't that a problem, that Horde/String.php never require_once'd in 
the whole file, but used String::convertChraset in 
http://cvs.horde.org/co.php/ingo/lib/Script/sieve.php?r=1.79 ??



Thanks,

Gergely

Great work! Tweaked and committed, thanks.

Sorry for being late.



In return, the fix is a lot more elegant now (to be precise: now it is 
correct ;) ).

And it also fixes another issue, ingo serialized accented vacation 
data wrong if database environment was iso-8859-2, while user 
environment was utf-8 (for the string length it substituted the length 
of the utf-8 string, but it wrote iso-8859-2 string as the string).   
Fortunately, the same problem did at decoding, so it only bugged the 
user if he (after saving a vacation string) changed the browser 
charset from utf-8 to iso-8859-2.



If something is not clear, ask!



Thanks for your patience,

Gergely

How's this going?

Please note, that I haven't ever received the message, which asks for 
clarifications about the patch.  I only received an email notification 
about the last comment an hour ago.



If everything goes well, I can provide a clear patch, with horde 
functions and with a configurable workaround for buggy sieve 
implementation at the weekend.  I will also try to notify the cyrus 
developers again, but my previous news article was lost by gmane...



Gergely

Looks like we have to fix this ourselves.

Please create a patch that uses Horde's conversion and encoding 
methods like MIME::encode() and String::convertCharset().

1. If the browser charset is not utf-8, than characters with accents 
don't get into cyrus sieve, since the specification says, that they 
must be UTF-8 encoded in sieve scripts.

Proposal:

  - send the form with accept-charset="utf-8" (my solution do this)

OR

  - encode the returned information from NLS::getCharset() to utf-8

Affected input fields: subject, body.



2. If the header contains accents, it is not enough to encode it to 
utf-8, but we have

to mime encode it, because of buggy mail servers (like Cyrus Sieve), 
which doesn't do this themselves.

Affected input fields: subject, from. (my patch only fix the problem 
on the subject field).