Hi
Joost, thanks for the reply (oddly it didn't arrive as gmail, just happened to see it here today).
There's no mention of GPano data in the Adobe XMP document you cite, so I think that at present
https://developers.google.com/streetview/spherical-metadata is the only useful/credible document.
It is definitely not generally true in XML that a name can be used either as an attribute or as the name of a content tag: for example, the 'toy' case:
<employee id=7543 font=arial>
<name> Fred bloggs <name>
<id>0573065</id>
<department> Sales </department>
</employee>
Here the attribute id is probably record number in database and the content id is probably the employee's serial number, etc. In general in SGML and XML, attributes should not affect the data content of a tag, though designers often get that wrong -- GPX being a classic example.
I'm also sure (unless it has changed in past couple of months) that Google Street View, Facebook, etc., expect the GPano: data to be in content (not attributes), as described in the document above.
You mentioned: "That image is from an old version of PTGui; since a few versions PTGui
is using Adobe's XMP toolkit for writing out the XMP data, so the format
may have changed. The XMP toolkit is probably the best way to parse XMP
data in all possible formats."
Good news: I just checked a spherical I made using PTGui a few months ago, and indeed the GPano data are content not attributes. So I don't think you need to do anything :-). I will drop the support for GPano in attributes maybe in next PanGazer release.
I doubt that I'd use the XMP toolkit because my own XML parser is extremely fast and the XMP processing that I need is almost trivial given the XML parse tree (about 120 lines of code, + defined constants).
Mike