IPTC verus XMP -- with a side of cars

[calberga]

I'm asking, here, for informed opinion as to WHAT I should do, not (yet at least) how to code it.

In tagging my images (.jpg) I have been using IPTC tags, with some custom tags created in the .ExifTool_config file.

These are mostly to provide a larger number of "location" tags, "Area" for things like "Hudson River Valley", "Adirondacks" (you can see I'm a New Yorker) which don't correspond to any official political entity, or to distinguish "Town" and "Township", or for colloquially named parts of places such as "Little Italy", "Dumbo", Tribeca" in Manhattan.

I would like to be able to create sidecar files for my .jpg files, for backup if nothing else, but as I read the exiftool documents (remembering that I am nether a perl nor a C programmer, just LISP and assembler code) it seems as if I have to have XMP equivalents for my home-made tags.

Is this impression correct?  If so, will the custom tags be any less opace to other people in XMP than in IPTC?  I.e., should I simply convert everything to XMP?  If so, should I define my XMP iptcExt tags, with the new location tags added to XMP LocationDetails Struct, or should I just create them in their own schema.

Or should I just forget about XMP and sidecars and build my own database as a backup?

I'm not sure I can follow the various explanations of creating new XMP tags, but think I can.

[Hayo Baan]

Since they are your custom tags, no one else will be able to do anything with them. So you're basically free to put them anywhere you want. XMP is the best place, not only is it the future now, it is the most flexible and is UTF8 by default so no problems encoding weird characters. I would add them where they are most logical, or indeed in their own schema as that keeps them isolated.

[Phil Harvey]

IPTC is moving from IIM to XMP.  You should take a look at the new IPTC XMP tags here and here.

- Phil

[calberga]

Thank you both.   Guess it's time for some research.

I suppose a third alternative would be to make them all keywords, and stuff all of the data in the existing XMP Keyword tag.  Then they would be accessable to anyone, albeit at a extra level of overhead.