Invalid SKOS for vocabularies with GUIDs

Yeah, I found this issue also months ago during a bout of extending/rewriting the dev code to use pure SKOS (D7-read/write-compatible) instead of the earlier W3C-thesaurus RDF examples the earlier code was modelled on.
Anyway, I got caught up in a few days of research into the state of play with other example/samples and how they managed (or failed to manage) the distinction between XML-base + IDs vs full-GUID-about attributes. And then other weirdo shortcuts and inconsistencies. And then tried to think about how *sometimes* my GUIDs are local and sometimes they are imported, but most of the ones I am importing start off being local to somewhere, and maybe I should stake a namespace like PURL and .. ah, I can't remember much after that, but that's the basic reason I never resolved it!

Yeah, and then I had to look very hard at D7 vocabs and GUIDs and the GUID module (I prefer machine_name) and ... that too, I didn't see a clear path in. I've been making it up a little as I go along, and only nominally writing this extra (probably useful, but not definative) extra info when I output. In my experiments I;ve found readers that accept rdf:about, and others that only assume xml:id, and only more recently has my habit of writing Both started hitting validators that actually care.
If I was writing a validator, and both methods were given, I should only complain if both were given and differed. As it is, I was (attempting to) say the same thing twice in to equally valid ways. That is, if the xml:base directive is still working in the current code. I can't recall.

I'm on holiday, and not in a position to test ...
But I see where you are going with this patch, it's pretty readable.
I though that any call to taxonomy_xml_get_term_guid() would return a valid value, aven if it had to make it up on the spot, so I would have expected

+    // If a GUID is available, it should be used instead of the ID to avoid validation errors
+    $guid = taxonomy_xml_get_term_guid($term);    
+    if (isset($guid) && !empty($guid)) {

to never trigger.
I may be thinking of a different version of the code though.

I think I'd still like to get my original big conceptual problem sorted out though - should we use IDs *or* rdf:about - as a straight rule, rather than gues on a term by term basis.
I really would prefer IDs (plus an xml:base) because that's much more succinct, easier to read and transfer, and portable. Also MUCH tidier for internal references. But xml:base support was flaky in the various parsers I was using (several years ago) and rdf:about has the advantage of being really portable even when things are broken up into small fragments.

This calls for a hand-wavy discussion over some strong drink.
OR a pointer at several consistent examples that say "this is the best way to do this" from several different respectable sources - that actually agree.