Nushell core team meeting 2023-11-29

Attendees

Lessons from the survey for us
Input/output types
1. commands communicating with each other and not having the appropriate input_output_type existing at all. I wonder if there's a way to put every variety of i/o type avaiable and then remove the ones that don't belong. Just thought it may be easier than having ones missing.
2. input_output_types being searched from top to bottom and choosing the first one versus the "right" one. Last discussed with Sophia, going top to bottom may not be as accurate as matching. I'm not sure and don't know where this code is at exactly but it definitely needs to work better to prevent developers from re-ordering i/o types in rust just to get a solution working.
3. Example: 11137, 11134, there are others too
Stabilization philosphy
- offline notes: Stabilization policy considerations
- order of things we guarantee to be stable or implementation defined for 1.0
- how to manage unstable things after 1.0, breaking changes apart from errata
- Survey of the land:
  - rust forever 1.0 and nightly
  - Python
  - Linux kernel versioning

Already a lot of people using Nushell in "professional" environments
- at work
- on servers
discrepancy between folks mentioning they use dataframes and the folks mentioning they enable the dataframe feature
sampling (bias) towards rust users?

Darren expectation bias: we have broken ones, so can you trust the errors?
missing input/output type pairs may still be lurking
order of input output types seems to matter for resolution
narrowing down the problem can take some effort across multiple commands to figure it out
raison-d-etre:
- error when things are not composable
- ide/completion narrowing of suggestion