[geotk] Problem with marshalling metadata

[Martin Desruisseaux]

Hello Emmanuel

The new revision of ISO 19115 standard will published in 4 parts:

  * What we were used to call "ISO 19115" will become "ISO 19115-1".
    This is the part is about UML (in our case, Java interfaces).
  * The ISO 19115-2 extension will keep the same name.
  * What we were used to call "ISO 19139" will become "ISO 19115-3".
    This part is about XML representation of ISO 19115-1.
  * XML representation of ISO 19115-2 will be specified in ISO 19115-4.


Only ISO 19115-1 has been published. The 3 other parts have not yet been
published. This means that we do not yet have a XML standard for the new
metadata model. Consequently, only the Java interfaces are upgraded to
the new ISO model in Apache SIS. The XML representation stay unchanged.

When upgrading the Java API for the new ISO model, we have some new
methods and some "removed" methods. The "removed" methods are actually
deprecated - they are not really removed. We try to not change method
signature as much as possible, but there is a few cases where we can
hardly do otherwise. E.g. some Collection<ResponsibleParty> become
Collection<Responsibility>.

All deprecated methods have a replacement in the new ISO model - there
is no lost. However the new model can express more information than the
old one, so there is some potential lost when using the new model in
Java API, then marshalling to the old XML schema. However Apache SIS try
to emit a warning in such cases. We do not catch all data lost, but
hopefully the main ones.

Apache SIS implementation of deprecated methods delegate their work to
the new methods. So there should be no difference in using old or new
methods - the data should be stored in the same place.

In my understanding, the European INSPIRE program will continue to use
the old XML schema until the next INSPIRE metadata revision, which would
happen in about 10 years. This means that the deprecated metadata
methods in org.apache.sis.metadata.iso (or at least our capability to
(un)marshall according the old XML schema) may need to stay there for 10
years. So metadata generated by Apache SIS would hopefully stay
compatible with Geonetwork for a while.

About the schedule: I hope to finish metadata upgrade in 2 weeks. We
already incorporated the main changes, and I started reviewing for the
changes that we missed.

    Regards,

        Martin


Le 23/07/14 11:48, Emmanuel Blondel a écrit :
> Thanks Martin,
> At now, i'm working on moving to Apache SIS on a separate
> branch...i've still this pending issue of marshalling (cf other post).
>
> Ideally, i would have liked to merge this branch, but I could try to
> use Geotk snapshot, also i'm very interested in the process of
> upgrading ISO19115 version. Do you yet know when this upgrade and new
> Geotk milestone are planned?
>
> How is going to be updgraded ISO 19115: is there backward
> compatibility with the first version or not? in other words, if i move
> to 4.0-snapshot (and later, the new milestone), Do i have to move to
> the ISO 19115:2014? My concern on this, is that i use geotoolkit to
> generate metadata that i then publish in a Geonetwork instance..., and
> i don't know if the ISO19115:2014 might give problems with the
> metadata publication. What do you think?
>
> Emmanuel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/geotoolkit/attachments/20140724/a12d1088/attachment.html>