"Re: [xml] serialization to UTF-16LE"

[Kasimier Buchcik <kbuchcik 4commerce de>]

Hi,

Daniel Veillard wrote:
> On Mon, Nov 03, 2003 at 01:47:48PM +0100, Kasimier Buchcik wrote:
>
> > Kasimier Buchcik wrote:
> >
> > > I get different serialized results for text-nodes and attribute-values 
> > > if using "xmlSaveFormatFileTo" (note that I'm using values like "öäü?").
>
>
>   Well, right this looks strange but text node and attribute-values must have
> different serialization routines anyway (line-feed need to be escaped as
> character reference for example).

Yes, I'm aware of those special cases. I just wonder what to do with 
letters that could be serialized more human-readable.

> > > The resulting xml:
> > >
> > > <?xml version="1.0" encoding="UTF-16LE"?>
> > > <foo bar="&#xE4;&#xF6;&#xFC;">öäü?</foo>
> >
> > Sorry, it has to be:
> >
> > <?xml version="1.0" encoding="UTF-16LE"?>
> > <foo bar="&#xE4;&#xF6;&#xFC;&#x20AC;">öäü?</foo>
> >
> > > The attribute value is escaped, the text-nodes value not.
> > > Since I did not found anything in the specs that states, that attribute 
> > > values need to be excaped, although they *could* be serialized with the 
> > > stated encoding (UTF-16LE), I'm asking for help on information; did I 
> > > overlook something in the specs, or if this is a not intended bahaviour?
> > > (Maby some context could be usefull: I'm just trying to implement (in 
> > > Delphi) the w3c's saveToString method of the DOMSerializer interface).
>
>
>   Well, how is that really a problem ? The serialization is different, 
> but have to be different anyway. I don't see why this would be not
> compliant or a problem for DOM implementation.
>  
> Daniel

Hmm, maby this would not be an issue of compliance, but of readability. 
Why using XML with special encodings if attribute values are escaped and 
consequently made non human-readable? Ironically one could say that 
values of text nodes could be escaped as well, but at that moment 
encoding would begin to make no sense.
However, do you see a workaround for this inconvenience? Which 
characters *need* to be escaped and which not? Can I hook in, somewhere 
in the lib, to adjust the serialization of the values of attributes?

Thanks,

Kasimier