@dakejahl I've always disliked the covariance matrices, but there is a significant cost to changing to Fix3 as we'd need to send both for many years, and we'd pay a flash cost of supporting both. The convention of mapping the covariance to length 6 and extracting the 3 ac curacies from that does work well, even though it is ugly
I think we'd need a stronger reason to go to a Fix3 than this, sorry

@tridge Understood. The primary motivation is quality metrics. Would you be open to another auxiliary gnss message for noise, jamming, differential corrections age, etc?