Been playing some more with flow and so far all continues to go mostly well (except for my open question on continuation ids and how to not pollute the cache with them?).
Up to now, in order to do validation, we've been running a transformer on the backside of a POST and invoking some complicated conditional logic in an endElement handler depending on the results of a Schematron transform farther up the pipeline. This works ok but is 1) hard to maintain, 2) impossible to debug, 3) completely opaque. I'd like to migrate this to a validation function in our flow script but don't quite understand enough about the Cocoon internals to see how to do this. Looking at the WebDav block suggests I could go that way, with some caveats: would it be possible to take the output of this transform as input to some subsequent pipeline and keep everything in memory? I really don't want to manage temporary file names and locking and I'd like to avoid any extra overhead. I take it that the Cocoon ModifiableSource is deprecated in favor of the Excalibur version? Is there any existing in-memory writable source available for reuse? Any hints on how to do this? Any documentation on the Excalibur stuff (where do I find it?) The other way to go is suggested from the woody samples and the binding example. However, it seems possible to just use a Cocoon pipeline on the request data (aggregated with our validation templates as created from our metadata) to feed the Schematron transform directly in the pipeline and not run the transform as part of the flow script? Any reason to head this direction? Peter Hunsberger
