Sept 20, 2022 Community Call

[expede]

Sign up: https://lu.ma/ipvm

Suggested Agenda

Let us know in the comments what you'd like to discuss! Questions about using UCAN, use cases in the wild, short demos, etc.

• @expede to more clearly delineate IPVM and Autocodec
• @softwareplumber describe ABI ideas related to XPath, XSLT, etc
• @RangerMauve exceptions, codecs, ADLs, byte loading
  • Quick @expede thoughts: exceptions are effects, go in job receipts
  • WASI article

Video

Chat Log

Notes

[RangerMauve]

It'd be cool if we could talk about what sorts of ABIs we'd like to use.

Personally, I think it would be prudent to settle on ABIs between wasm and the host for the following (or to see if other folks are even into this approach):

• Memory management. How can we tell the host that the wasm module no longer needs a pointer
• What should exceptions look like (what exists)
• Codecs:
  • Providing bytes to a WASM module
  • APIs for the WASM module to construct an IPLD tree (new map, add this property, add a sub-map, etc)
  • Can this be done lazily, does it have to be all at once?
  • Do we use a similar pattern to NodeBuilder
• ADLs
  • Use same APIs that WASM codecs use to represent a new Node
  • Use mirrored APIs for WASM to get a Node represented in the host
  • API for a WASM module to get a Node pointer by it's CID
• Spec for how WASM modules get loaded from IPFS
  • UnixFS for modules larger than a block?
  • Support ADLs like FBL as alternative to UnixFS?
  • What languages are folks willing to start implementing stuff in?

[RangerMauve]

Also would be cool to talk a bit about the WASI stuff Adin linked to

[RangerMauve]

https://devclass.com/2022/09/08/assemblyscript-project-wasi-damages-open-standards-and-the-web/

[nicobao]

Hi,
Thank you the call earlier, it was very interesting to listen.
I have a question, web assembly is promising and exciting, but docker - despite its limitations - is already a mature way to scale computing over a bunch of machines.
Apart from bacalhau, is IPVM working group also working towards this direction?
On the FVM side - is it something that is being thought about as well? Or is there some technical limitations here that I am unaware of?
We could very well design an IPLD Data Model that would map the docker requirements and then run that externally, no?