Design Changes since Original (2015)

as of 2022/03, the design has not changed much, but sr3’s implementation is totally different from v2. Design changes:

• explicit core application support for reports were ripped out as they were never used, can easily be re-inserted as callbacks and conventions. (re-inserted in 2023.)

• nobody used segmented files, and they were very complicated, but everyone finds them fantastic in theory. Need to re-implement post sr3 re-factor. (re-implemented in 2023 and 2024)

• mirroring was a use case we had to address, had to add metadata, and evolve a bit. (

• the cluster routing concepts have been removed (cluster_from, cluster_to, etc…) it got in the analysts’ way more than it helped. Dead easy to implement using flow callbacks, if we ever want them back.

• the Flow algorithm emerged as a unifying concept for all the different components originally envisaged. In early work on v2, we didn’t know if all the components would work the same, so they were written separately from scratch. A lot of copy/paste code among entry points.