There’s another, albeit likely much smaller issue, which is that Roon’s reference platform (ROCK) is Linux-based. Dirac, that I know of, doesn’t currently have publicly available Linux binaries. My understanding is that they exist for ARM platforms, but who knows how much (or little) work it’d take on their end to make that work, or work well, on Linux x86. Going forward, the licensing question is obviously the bigger one, moreso with Roon possibly getting remote streaming capability at some point in the future. I’ll take Dirac’s announcement of “modules” (and the “car” module being one of the first examples Flavio gave) as a sign it’s something Dirac has thought about.
This said, anything that isn’t $800 a seat, which is essentially what Dirac costs right now if you factor in the dedicated hardware to run it (be it from MiniDSP or rolling an endpoint with a commercial OS (and all the compromises that entails)) would be progress. Something else I can’t help but wondering about when it comes to room correction in general is who gets to deal with customer support. Dirac have done a truly remarkable job at simplifying the process, but it still isn’t completely idiot-proof, and that’s something that possibly has a not-inconsequential cost.
Generally, my personal feeling is that it’d be more constructive for us, as users, to lobby for “any reasonably user-friendly, known good, advanced room correction” rather than “Brand X” room correction, and trust RoonLabs to find the best compromise for everyone. I, for one, don’t care if room correction is done by Dirac (though I’m impatient to hear how good the new algos sound), Trinnov, or something else (maybe there’s a way for Roon to licence HAF’s algos and make whatever manual tuning is going on there user-facing, for example), just as long as it’s good, and isn’t as involved and obtuse as REW/RePhase (no matter how great @Magnus’s and @SwissBear’s tutorials are 
).