Interesting to see FaaS and Serverless compute start to diverge.
Strongly believe the latter needs to eliminate all routing, placement, scaling, management, packing, and patching responsibilities from your domain while meeting your availability, latency, and scale needs. Doing any of that yourself is work you shouldn’t do.
FaaS is one of (and IMHO, the best suited) programming model to best realize the Serverless model, as a way to put together serviceful architectures. But by itself, it won’t give you the cost and agility benefits you have come to associate with Serverless stuff.
Totally hear and understand customers wanting to use the FaaS programming model on compute stacks they own and operate, and for wanting to delegate some tasks (instance management) and not others (routing/placement).
Lots of cool and interesting work in making that easier, and it will always be a continuum (I mean people still build datacenters). Just don’t confuse it with eliminating all the work that going Serverless eliminates.
