From: Simon P. <si...@dp...> - 2007-11-22 02:02:21
|
Michael Gannon wrote: > Hi Simon, > > That all sounds great, the part that I was particularly eluding to was > addressed by you with this: > > "first against the XSD (with a more sophisticated error reporting mechanism > still to be worked out)" > This is the part where your input is needed :-) Do you want to be able to plug in your own validation tool here (eg. xmllint) which returns an html report? If no plug in is configured then the default behaviour is the current cryptic stuff produced by the GN parser? > While I appreciate the space the Schematron works in there is likely to be > more errors, and more cryptic errors thrown from XSD's than the Schematron. > I think schematron can be used for much more than just validation as XML people understand it - think of the checks on content that could be made for an organisational metadata profile using schematron rules. So whilst I agree that the cryptic ones (including the ones that are difficult to pin down usefully) are going to come from the XSD validation, it is schematron that will probably be the most prolific once it is understood how useful it is. Anyway - this is kind of a side issue - so sorry for diverting off here. > Just to throw my 2cents in I would have through fixing the reporting of XSD > errors would have come well before implementing Schematron (although I know > where you are coming form with the ANZLIC requirements) giving you a stable > place to work from in implementing further validation be it XSLT, SCH etc. > I think from our experience in the bluenet area where entry is via the geonetwork editor and controlled by templates, there aren't that many XSD errors (when the editor understand the schema correctly! :-)) and of those many could probably be caught and reported by schematron more usefully at present - so it depends on where you're coming from. This would be different for harvested records and XML imported from other places though - particularly in organisations where ISO metadata is created from a number of different internal systems and then imported into GN so I take the point - see above. > That's just my 2cents but it looks like we are getting there. If you had to > put a time on it, or perhaps a version might be more appropriate, when would > you hope to see some of this work fixing the error reporting with XSD's > getting into the stable builds? > > I don't think offering a config option to run an external validator like xmllint that produced an html report would be hard (thats what schematron does now) and I think thats the option I'll put in the 2.1 test I'm building now - but there is more out there than xmllint (don't even know if xmllint is available for windoze people) so more input is required... Cheers, Simon |