Repository navigation
Technical meeeting 2026-11-18 #4878
liamfallon
started this conversation in
Community meetings
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Agenda
Attendance
@liamfallon
@ahmadu2305
@AndrasMNokia
@CsatariGergely
@zaadnsu
@aravindtga
@Catalin-Stratulat-Ericsson
@efiacor
@JamesMcDermott
@mozesl-nokia
@rendre-greyling
@timop62
If you are not on the attendance list add your GitHub handle to a comment under the discussion
Minutes
@liamfallon chaired the meeting.
Priority prs
Security check
Kpt release WoW
2026-08-12:
2026-08-19:
2026-08-26:
2026-09-02:
2026-02-09 to 2026-09-16: No progress (@efiacor on leave)
2026-23-09: Document release dependency chain and remove replace directive friction between api and kpt modules #4769
2026-09-30: Does go have mech to mark api fields as experimental/open to change
2026-10-14: Github action could streamline this, @efiacor to raise an issue
2026-10-21: No update
2026-10-28:
2026-11-04:
2026-11-11:
2026-11-18:
Local git cache
2026-10-28:
2026-11-04:
2026-11-11
2026-11-18
CR Cache and Caches
2026-10-28
2026-11-04:
2026-11-11:
2026-11-18:
Should we ask community members to participate on the meetings?
2026-11-18:
WASM support
2026-11-18:
The way it is implemented in kpt (or rather, the functions themselves) has been mostly unsupported since golang ~1.21 and relies on experimental go features. We should decide how to move forward now that there are compiler changes breaking our currently used lib (see #4866).
We can keep the current “javascripty” function implementations working by replacing wasmtime with wazero, but we essentially need to maintain our own “shim” that connects syscall/js with wazero . (I am using these quoted terminologies very loosely)
We can drop WASM support entirely, not sure if there are any users out there.
We can move to supporting WASM in some other way, like WASI, which could just be few lines to run using wazero in kpt, no “shim” needed. In this case, the functions have to be built for WASI, which would be a major change there.
(Feel free to correct me if anything I wrote is wrong!)
We should drop WASM support because
It's not really supported very well right now
It would be costly to implement correctly
We have no users (that we know of)
The current implementation, even if it's fixed, could be a security risk (Attack surface)
Agreed: Let's drop WASM support. @mozesl-nokia to raise an epic for this work
EPIC: Remove WASM support #4881
Non-KRM Resoruces
Would we want to support running functions on non-KRM resources? Meaning it would be “raw” yaml without kind/apiVersion fields. My primary use case would be config files with setter annotations that would later be embedded into a ConfigMap. The alternative would be to support setters on embedded yamls in ConfigMap data.
2026-11-18:
All reactions