MV3 Beta 3.0.0.1114: regression in Brave UX
lidel opened this issue · 5 comments
@whizzzkid I've found two bugs in Brave:
Problem in MV3 Beta 3.0.0.1114
When built-in node is used in Brave:
Companion disables "My Node" button:
And redirect happens to http:// URI instead of native ipfs:// and ipns:// :
Expected behavior
- Opening WebUI after clicking "My node" should work.
- When user is redirected to native URI when built-in Brave node type is used:
@lidel I think that's because the redirect uri being generated for brave is not ipfs:// or ipns:// hence the redirect happens to http:// the pattens are strictly based on the redirect-uri
I think I can have fix for this.
Issue 1: My Node Links, it seems to work for me. I'm not sure what broke or fixed it.
Issue 2: not redirecting to native ipns:// or ipfs://
The simple fix I was thinking was reverting to the old blocking API like in FF, but that's not the case here, brave does not implement the blocking api in MV3, yet, the specific rules to native protocols seems to be missing, It will need some more work. I can work on it as an enhancement after initial MV3? or will that be a blocker?
@whizzzkid thanks for looking into this. It feels like a blocker for publishing to stable channel, some details, and ideas below.
On why we may want to fix this before mV3 ships
Brave uses the package we publish to Chrome Web Store, so by shipping MV3 without fixing this, we introduce an unnecessary regression in Brave, which aims to be the state of the art for native handlers like ipfs://.
We would be undoing years of progress: we have a browser which supports native handler, but we make it act like regular Google Chrome or Firefox. We also alienate existing users who had native protocol for years now, and if someone was using something like wallet over it, they may lose access to it unless they type ipfs://cid manually.
Not the end of the world, but imo easier to fix now,
rather than to deal with fallout and responding to issues here and in brave.
Potential implementations paths
I don't know if/when a blocking API in MV3 will happen in Brave. We could ask Brave to hardcode and allow blocking API from MV2 for Companion, but no ETA on this. Imo not worth asking, as it is not a stable foundation we should rely upon going forward anyway.
See if we can use declarative API.
If not, perhaps we could fix the address bar in Brave in less invasive ways.
One idea, introduce an additional listener in Brave that looks at chrome.webNavigation events of the top level documents (URL in address bar), and when a resource with http:// URL of the local gateway is loaded, update the tab and replace it with ipfs:// equivalent (so only ipfs:// shows in browsing history).
Bit hacky, but reload will happen so fast end user will not notice. The benefit of this approach is future-proof, and will work no matter how broken declarative API in MV3 is :)