Are language tags a moving target?

Recently I asked the question List of RDF literal language tags and learned that the definitive list is ISO 639-2. My original motivator for asking that question is to figure out what language I was looking at when I downloaded the Titles dataset for the "ga" language tag from Dbpedia. I had looked at some other resource on the web that said the "ga" tag was for Irish, which seemed to jive. But then looking at ISO-639-2 I saw that ISO-639-1 tag for Irish was "ga" and for ISO-639-2 it is now "gle" and ISO-639-2 uses the tag "ga" for the spoken language Ga, a tribal African language.

That's a pretty big switch. Since I find it unlikely dbpedia would publish titles for a lesser(?) spoken tribal language in the small set of languages it does publish for, over Irish, I'm inclined to think DbPedia is using the ISO 639-1 language tags. Which makes me think if a somewhat authoritative dataset is using the outdated spec, who else has been tripped up by this moving target, and what issues will it create for Linked Data?

But my question is: are the language tags still being defined? And, if so, does this present a problem for Semantic Web adoptance?

oh and my favorite tag is "tlh" :-)

ISO codes tend to be pretty stable. With country codes they have made a commitment that when a code is retired (which for example can happen when a country changes its name, or splits up into smaller countries) it won't be reused for 50 years. I'd expect them to assign language codes in the same spirit.

And "jbo" beats "tlh" any day of the week.