#"platform modules" design

1 messages · Page 1 of 1 (latest)

inland turtle
#

@steel zealot @snow wraith my brain is mush this week with jetlag & heat... I am struggling to form a design opinion but really want to make progress on this, since it affects platform teams so much

steel zealot
#

anything specifically you need a hand with? I feel like the discussion we had, I'm happy with the outcome of that - or at least, happy for us to prototype that (since it split the daggerverse into two, probably worth looping more folks into discussion, if it works)

inland turtle
steel zealot
#

(my brain's a bit fuzzy today, not feeling at 100% 😢)

but:

  1. I quite like "toolchain". "template" works, but to me, implies generation once, instead of kinda constantly updated. maybe even just "upstream" works as well?
  2. do we need a trust system? i'm a bit confused by this, what's the problem we're trying to solve? i'm not convinced we need this?
  3. no super strong preferences, but i like the idea of using the same install command - there's no ambiguity if the "platform module" can't be installed as a regular module...
#

trust wise, I think prompts are useful for when we need to get the users consent during an operation - we don't know all the details before (since it depends on execution), so llm API is perfect for that.
but in install, it's very procedural and straightforward - if we are worried about a security issue (not sure what that is?), then I think a warning message or even a flag passed to dagger install would be good (maybe that's a good reason to have a separate command)

inland turtle
steel zealot
inland turtle