Harmonize POST API vs jupyverse?

[cben]

I was trying to make jpterm work with this server, and discovered the HTTP API adopted here vs. the one adopted by jupyverse are different.

• jupyter-server-nbmodel sends nested
  { code, metadata: { cell_id, document_id, record_timing } }.
  Parsed here then here; code snippet is mandatory.

  There is also additional GET .../requests/<taskid> API for polling,
  plus POST .../input API for input().

• jupyverse client sends sends flat { cell_id, document_id }.
  Server looks up the code itself from its Y model.

The nested/flat is cosmetic but the 2 code approaches are an interesting tradeoff! 🤔

• race condition: In collaborative editing, if I Shift+Enter a cell you're editing but my client didn't sync it yet, jupyverse risks executing new code different from what old code I intended to run?
  OTOH this extension will execute precisely what I saw — but then notebook syncs with new code, risking showing results that don't match the code? Will final notebook mark cell as out-of-date in such scenario?

Dunno what are the long-term plans, but figured I'll document the difference.

[cben]

and I see README here already comments on this:

The code snippet is always send in the body of the POST /api/kernels/<id>/execute request to avoid document model discrepancy; the document on the backend is only eventually identical with the frontends (document updates are not instantaneous).

[krassowski]

Will final notebook mark cell as out-of-date in such scenario?

I think this would be the desired behaviour. I am not sure if this works as of today, but the orange dirty indicator should in principle show up then.