I got constant strings which I typed in the script with accented chars UTF8-encoded.

If your source code is UTF-8, use use utf8;.

>perl -le"binmode STDOUT, ':encoding(iso-latin-1)'; print qq{print len
+gth '\x85'}" | perl -l
1   Good

>perl -le"binmode STDOUT, ':utf8'; print qq{print length '\x85'}" | pe
+rl -l
2   BAD!

>perl -le"binmode STDOUT, ':utf8'; print qq{use utf8; print length '\x
+85'}" | perl -l
1   Good
[download]

Is it possible to "switch" a script in such a way that ALL outputs get encoded with respect to some locale setting we can read from the system?

#!/usr/bin/perl

# This source file is encoded using UTF-8.
use utf8;

use strict;
use warnings;

# Use locale-dependent encoding for STDIO.
use open ':std', ':locale';

# Use locale-dependent encoding (by default)
# for all files opened in this scope.
# Unfortunately, <> ignores this directive.
use open IO => ':locale';

...
[download]

It might makes more sense to use a known encoding for files, though.

use open IO => ':encoding(UTF-8)';
[download]

There's also File::BOM in case you want to accepts UTF-8 (and -16le and -16be) while giving allowing a fallback to another encoding such as iso-latin-1.