I'd understood distributed transactions to be a vendor-specific thing. I'd never have realized I could have IMS/DB, DB2, and MSSQL all in the same txn.
Candidly this is not my area of expertise but I always thought it was multi-vendor. Lacking first-hand knowledge, a curious quick web search does imply that at least the Open/XA standard was supposed to be multi-vendor:
The Open Group has defined an industry-standard model for transactional work that allows changes made against unrelated resources to be part of a single global transaction. An example of this is changes to databases that are provided by two separate vendors. This model is called the X/Open Distributed Transaction Processing model.
Is Microsoft's MSDTC more single-vendor? Even that seems to support multi-vendor databases through its support of Open/XA:
When the DTC acts as an XA-compliant transaction manager, Oracle, IBM DB/2, Sybase, Informix, and other XA-compliant resource managers can participate in transactions that the DTC controls.
Diving into the history of which distributed transaction manager was the first to support multi-vendor database transactions is left as an exercise for the reader...
I suspect industry knowledge/awareness of this stuff waned as web-generated transactions boomed and cloud platform vendors like AWS sold "eventual consistency" to developers and through them back to management and end users (like myself!) who think "Oh, that's weird, why didn't X update? Let me just refresh my browser... oh, there it is".
I've met a lot of people (typically open source enthusiasts) who say that distributed transactions are a dumpster fire and other people (Microsoft/enterprise partisans) who think they are great.
My go-to for a critique of distributed transactions is "Life beyond Distributed Transactions:
an Apostate’s Opinion"[1] which says in the abstract "performance costs and fragility make
them impractical". That was 2007, though.
