Absolutely. There *is* a globally unique, resolvable identifier for publications: Digital Object Identifiers (DOI). It's a distributed, hierarchical system: organisations that publish works are assigned a root DOI and run a DOI server. Furthermore, DOIs can be assigned not simply to a publication, but to individual items in it, drilling down as far as you like.

My impression is that DOIs mainly apply to things like journal articles: things that are being published today by publishers with an online presence. Many of our references here are CSIRO publications - I wonder if CSIRO publishing hosts a DOI resolver?

A problem is that while given a DOI you can easily get the metadata for it, discovering the DOI for a publication is ... an effort. I imagine that another problem is homonyms: if two different publishers are producing editions of System Naturae,  there will be different DOIs for it.

Another issue is that if you want to cite a page of a journal, or better - a particular figure in the journal, and the publisher does not publish a DOI for that particular thing ... where do you go from there? You could host your own DOI server and assign your own DOIs .... but that rather defeats the purpose. There's no getting away from having a "microreference" field, allowing a human to fill in the gaps in the identifiers with free text.

See: 

__________

The list of formally registered URN namespaces is here:

The ones potentially of interest - as far as I can see - with respect to serving as a globally unique citations are

These probably only go down to the "volume" level, but that's usually all you need for books.

There are also urn namespaces for things like articles published by news organisations, audiovisual files, if you want to cite them.

See: