Re:Interoperable Serendipity
by btbuildem on Thursday October 14, 2021 @11:47AM
(#28865898)
The crux of the problem lies in the semantic mapping (lenses / shapes in the article) and the transformations potentially required to map data that is not 1:1
Standards change / evolve, and for each "lens" you have an input and and output schema to keep track of (and potentially, the transformation logic associated with it).
This implies the existence of a transformation layer, capable of both running the logic and accessing the data - which prompts the question "where will this live".
In an adversarial interop scenario, I would expect the "adversary" (the side making the effort) to be hosting this. In a system built for interop from ground-up (yes, pipe dreams) it could be a dedicated subsystem that also enforces access control on the data.
Standards change / evolve, and for each "lens" you have an input and and output schema to keep track of (and potentially, the transformation logic associated with it).
This implies the existence of a transformation layer, capable of both running the logic and accessing the data - which prompts the question "where will this live".
In an adversarial interop scenario, I would expect the "adversary" (the side making the effort) to be hosting this. In a system built for interop from ground-up (yes, pipe dreams) it could be a dedicated subsystem that also enforces access control on the data.
It's definitely an interesting problem!