Subj : Re: UTF vs specific
To   : Karel Kral
From : Eugene Subbotin
Date : Tue Sep 15 2026 02:06 pm

Hello Karel!

Tuesday September 15 2026 11:48, you wrote to me:

 KK> OK, started step by step. OS is UTF-8.

 KK> GROUP C
 KK> XLATPATH /home/fido/golded/xlat
 KK> XLATCHARSET CP895 LATIN-2                       kam_il2.chs
 KK> XLATCHARSET LATIN-2 CP895                       il2_kam.chs
 KK> XLATIMPORT CP895
 KK> XLATEXPORT CP895
 KK> XLATLOCALSET UTF-8
 KK> XlatConfigSet LATIN-2
 KK> ENDGROUP

 KK> In that group "CP895" is compulsory. Does not work. Kludge is OK, what
 KK> I see is rubish.

 KK> (even if I try LANG=cs_CZ.ISO8859-2 - or even if I put original
 KK> command - luit -encoding 'ISO-8859-2' ./golded -f I had before).

The config cannot work as written, and the reason is simple. GoldED+ 2.0.0 asks 
iconv first for every conversion.
A *.CHS table is used only when iconv has no converter for the pair, and it 
looks for an XLATCHARSET line naming exactly the pair it needs. In a UTF-8 
session the pair for reading is CP895 -> UTF-8. Your tables are CP895 <-> 
LATIN-2, so nothing ever uses them: the log says "No recoder from CP895 to 
UTF-8; falling back to charset tables", finds none, and shows the text 
unconverted. That is the rubbish.

A few more things in what you posted. XLATCHARSET, XLATPATH, XLATLOCALSET and 
XLATCONFIGSET are global keywords.
Inside a GROUP they are read as global settings at that point of the file, and 
XLATLOCALSET/XLATCONFIGSET must be the first lines of golded.cfg, before 
anything non-ASCII, or everything above them is read in another charset.
Only XLATIMPORT and XLATEXPORT belong in a GROUP. XLATCONFIGSET LATIN-2 says 
golded.cfg itself is LATIN-2 - if the file is UTF-8, say UTF-8.
And luit or LANG=cs_CZ.ISO8859-2 translate the terminal while XLATLOCALSET 
UTF-8 tells GoldED+ the terminal is UTF-8. The two undo each other, use one or 
the other.

There are five tables in GoldED+'s CHS: kam_u8.chs (CP895 -> UTF-8), 
kam_852.chs and 852_kam.chs (CP895 <-> CP852), kam_il2.chs and il2_kam.chs 
(CP895 <-> LATIN-2). Letters follow the Kamenický layout, 0xB0-0xFF are the 
CP437 box characters as in CP895:

https://drive.google.com/file/d/1WrQYRsw7p4q5yfceqgISClMHl_RLHuEJ/view?usp=sharing

For your UTF-8 session put at the top of golded.cfg

XLATLOCALSET UTF-8
XLATCONFIGSET UTF-8
XLATPATH /home/fido/golded/xlat
XLATCHARSET CP895 UTF-8 kam_u8.chs
XLATCHARSET CP895 LATIN-2 kam_il2.chs
XLATCHARSET LATIN-2 CP895 il2_kam.chs
XLATCHARSET CP852 CP895 852_kam.chs
XLATCHARSET CP895 CP852 kam_852.chs
XLATCHARSETALIAS CP895 KAMENICKY KEYBCS2 NEC-867 DOS-895

And in the group:

GROUP C
XLATIMPORT CP895
XLATEXPORT CP852
ENDGROUP

CP895 messages then read correctly and what you write goes out as CP852 with a 
proper CHRS kludge. Writing CP895 from a UTF-8 session is not possible with a 
table now (I'll add reverse-encoding later): a CHS table maps one byte to up to 
three, so it can turn a CP895 byte into a UTF-8 character but cannot read a 
UTF-8 character, which is several bytes. With XLATEXPORT CP895 in a UTF-8 
session GoldED+ finds no converter and writes UTF-8 with "CHRS: UTF-8 4".

If an area REALLY has to stay CP895, run an 8-bit session instead:

XLATLOCALSET LATIN-2
XLATCONFIGSET LATIN-2
XLATPATH /home/fido/golded/xlat
XLATCHARSET CP895 UTF-8 kam_u8.chs
XLATCHARSET CP895 LATIN-2 kam_il2.chs
XLATCHARSET LATIN-2 CP895 il2_kam.chs
XLATCHARSET CP852 CP895 852_kam.chs
XLATCHARSET CP895 CP852 kam_852.chs
XLATCHARSETALIAS CP895 KAMENICKY KEYBCS2 NEC-867 DOS-895

Then XLATIMPORT and XLATEXPORT CP895 in the group:

GROUP C
XLATIMPORT CP895
XLATEXPORT CP895

And a any terminal showing ISO-8859-2 (or luit). Then both directions go 
through the tables and a written message carries "CHRS: CP895 2". CP852 works 
the same way with the 852 tables.

But PLEASE do not post CP895 messages to Fidonet!
Neither the Fidonet standard FTS-5003 nor IANA recognizes this encoding.
Nobody else has a table for it - iconv, Windows and OS/2 do not know the 
charset - so every other reader, GoldED+ without tables included, shows a CP895 
message as rubbish, exactly what you saw.

Czech and Slovak messages in Fidonet are CP852 written on the DOS systems 
("CHRS: CP852 2") or LATIN-2 (AKA ISO 8859-2) written on any other OS ("CHRS: 
LATIN-2 2"), and GoldED+ 2.0.0 reads and writes both without any tables:

GROUP C
XLATIMPORT LATIN-2
XLATEXPORT LATIN-2

The tables are for reading old CP895 mail. So, please write only in CP852, 
LATIN-2 or UTF-8.

Eugene

.... It's full of stars!
--- GoldED+/BSD 2.0.0-b20260912 (NetBSD 11.0 Intel Xeon Icelake)
 * Origin: FireFox Station (2:5075/35)

.