barneycarroll
commented
3 years ago Following on from #10, the thinking here rests on deferring implementation complexity to the vague deus ex of Island. As argued there, the unicorn of 0-DOM-injection-sentinel is unproven and as far as conceivably proven reliant on flaky & expensive runtime inference at significant engineering cost and reduced reliability.
As it turns out, Island was in practice too abstract a unit to implement with any consistent guarantees: Mobile fulfils a more holistically coherent remit than Portal and achieves all the meaningful conceptual challenges of Island. Portal is ultimately a puerile achievement asking for colossal backing implementation — undesirable.
Current logic dictates that prior content is trashed and new content initialised from scratch on every draw (if the outlet appears before the corresponding inlet).