Am Dienstag, den 06.03.2007, 17:43 +0100 schrieb Hendrik Sattler: > Am Dienstag 06 März 2007 10:53 schrieb Daniel Leidert: > > > is there any reason why ISO-8859-1 is enforces and entities used for > > > everything that does not fit in? > > > > You should ask docbook-xsl upstream. > > > > > Why isn't the encoding of the XML file used? It has to fit anyway and > > > makes the output much more readable! > > > > Simply: It's impossible. Read > > http://www.sagehill.net/docbookxsl/OutputEncoding.html: <cite>[..] That > > is because the XML specification does not permit the encoding attribute > > to be a variable or parameter value [..]</cite> > > > > > I assume you agree that forcing ISO-8859-1 does not make much sense with > > > UTF-8 input (except for some countries). > > > > You can choose an UTF-8 encoding for output using the xsl:output > > template or the chunker.output.encoding parameter in a custom stylesheet > > or via the -m option of xmlto (see also > > http://www.sagehill.net/docbookxsl/OutputEncoding.html). > > > > IIRC there were reasons, why ISO-8859-1 was used. But I cannot remember > > atm. However, I don't see a bug here. Therefor I'm going to close this > > report now. Feel free to comment this decision or reopen it. > > ISO-8859-1 is not suitable for automatic processing. AFAIK they argued with > old browser that do not understand UTF-8. What a lame excuse, though :-(
I don't think, it's "lame". I really observed issues with HTML output and UTf-8 some time ago. But to be honest, I never examined that deeply. > The XHTML output has UTF-8, IIRC. Yes, that's the case. > BTW: same goes for manpage export. Here, man is definitely broken, though > (truncates manpages with UTF-8 even in an UTF-8 locale). I don't know, what you mean. Can you give an example? Maybe you just missed the man.charmap.use.subset parameter setting/explanation? > Working with custom stylesheets is possible but NOT the solution: just > imagine > that everyone puts out his own stylesheets just for a sane character set! Where would be the problem to simply import the default stylesheet and set xsl:output/chunker.output.encoding accordingly to your wishes? Many processors allow a simple param=value addition to their command line. I'm currently working on stdin-support for xmlto, so it will also be easier there to specify [..] -m <xsl:param name="foo" value="'bar'"/> [..] There is also an encoding guess in xmlto. It sets the chunker.output.encoding parameters accordingly to it's guess. We can improve this guess or it's functionality (like in Debian bug #327551), if you think, that can solve your problems. Maybe by giving the requested encoding via command-line? Or would the patch in bug #327551 solve your problem? > That's insane and just produces lots of duplicate stuff along with lots of > broken stuff :-( Why it should produce "broken" stuff? For me it currently looks like you are demonizing the customization layer. I have no idea why this should be a good idea. Regards, Daniel
