As I hinted at on fop-users, I'd like to look at this in a more general way. Nils raises a valid point. FOP should not complain so loudly about foreign properties. The same applies to elements. With my recent changes we got rid of complaints inside SVG sub-documents as they, too, can contain elements and attributes from arbitrary namespaces that FOP doesn't know anything about. Still, we get complaints from FOP if we have elements in a foreign namespace somewhere in the middle of FO content.
Nils' patch is fine as a simple non-final solution. Still, we need some kind of extension mechanism on attribute level at some point and hopefully one that can make use of the property features defined by XSL-FO (inheritance, functions etc.). I have some desire for extension properties on external-graphic and instream-foreign-object: cache hints, rendering hints for extensions. FOP should also stop to complain about certain foreign elements. The complaints are fine as a debugging aid, for example if someone makes a mistake with the namespace URI or the name of an element in a particular namespace. Here's where I got the idea that we could provide for a list of namespace URIs which are simply silently ignored (instead of having to write a FOP extension for each namespace). If someone has a reference to a namespace URI in his documents that FOP doesn't know about he could add that namespace URI to that list and FOP will fall silent over it. The same list could be used for handling foreign attributes. This way you still get important feedback if you've done anything wrong, but can tell FOP to shut up where necessary. WDYT? On 18.02.2006 10:21:36 Andreas L Delmelle wrote: > On Feb 18, 2006, at 07:45, Nils Meier wrote: > > Hi Nils, > > > I looked into this some more and it seems to me that > > if PropertyList would check attribute namespaces > > before trying to convert them this would be handled > > nicely (i figured out that fobj has a reference to > > the responsible element so accessing its namespace is > > easy). > > > > I might be going out on a limb - maybe logging unknown > > attributes is the right thing to do(?) But why would > > FOP even consider attributes out of a non-fo namespace? > > > > well, here's a patch that does the namespace check > > just FYI. Looking forward to hear what you guys think. > > Your point is definitely valid, and I think I'm going to apply this > change. If no other devs object to this, that is... > (Later we may add an 'else' to that 'if', to deal with possible > extension properties.) > > Thanks for the feedback! > > Cheers, > > Andreas Jeremias Maerki
