Re: problem with ISO-2022-KR encoder

[Ulrich Drepper <drepper at cygnus dot com>]

Bruno Haible <haible@ilog.fr> writes:

> The glibc-2.1.1 iconv ISO-2022-KR encoder puts an "Esc $ ) C" sequence
> only once, at the beginning of its output, not in every line.
> 
>[...] but RFC 1557 says it must appear once in every line
> containing SO characters (rationale: so that if some lines of the text get
> lost, the remaining are still recognizable as Korean).

Where do you read this?  The formal description in RFC 1557 is:

   body            = *e-line *1( designator *( e-line / h-line ))

   designator      = ESC "$" ")" "C"

   e-line          = *text CRLF

   h-line          = *text 1*( segment *text ) CRLF




   segment         = SO 1*(one-of-94 one-of-94) SI

(Closing parenthesis in `segment' added by me).


Of interest is the `body' line.  This is not a recursive rule.  And
since *1 in front of

	( designator *( e-line / h-line ))

means "zero or one time" there must be exactly one occurrence of the
designator and this must happen before the first use of SO.  Therefore
I think the glibc implementation is just fine and RFC 1557 does *not*
contradict Ken Lunde.

-- 
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------