At 05:32 AM 3/30/2005, Jeff Trawick wrote: >On Tue, 29 Mar 2005 17:41:31 -0600, William A. Rowe, Jr. ><[EMAIL PROTECTED]> wrote: > >> I would also propose we drop apr_xlate and mod_charset_lite if the >> only way to support this is GNU iconv, which incompatible with the >> ASL. > >Various platforms have non-GNU iconv() which is suitiable for use by >this apr and httpd code. Most of my original testing was done with >native, non-GNU iconv() on z/OS and Solaris.
Yes. The issue is, there (was) no maintainer for that code for some years. It was rather DOA. There has been traction by Alexander (bland at freebsd) to re-support the Konstantin code. The newlib folk ported Konstantin's 2.0 code as well. Konstantin's mirror is now down and his code is no longer (directly) available. So we end up with three groups using the same implementation, but no common ground yet. I hope this will change (very shortly.) >Just curious: Suppose we're talking about Linux, which provides GNU >strcpy() and GNU iconv(). What is the difference with using one vs. >the other? Is your concern limited to those platforms which don't >provide iconv() out of the box, and where users may feel inclined to >install GNU iconv() in order to use the APR and/or Apache feature? Nothing. strcpy() is part of ANSI, C99 etc so we trust it's there. iconv() is a new gcc addition (at least, to clib) and while good, we don't have a non-GNU solution. We don't object to users plugging in GNU libiconv. Nor should we. But we won't distribute it, and should not require it. Therefore we have an issue when no alternative exists. As the introduction of the Euro symbol demonstrates, code pages -do- change over time, so an orphaned solution is no solution. Bill
