ci: rework publish job to use bundlr+ARNS to host code on Arweave subdomain

[TillaTheHun0]

Have tried swapping to a fork, eggy, but am running into some issues. Will need to investigate further.

[TillaTheHun0]

Waiting on a response on this issue deno911/eggs#8

[TillaTheHun0]

Ultimately, we just want our code up on Arweave.

Another approach may be to just use bundlr to upload onto Arweave and then use g8way.io and Path Manifests to import directly from Arweave with no proxy service like nest.land being required.

[twilson63]

Yep, happy to help there

[TillaTheHun0]

Currently with nest.land we import from https://x.nest.land/{module_name}@{version}/mod.js which redirects to an Arweave transaction like https://arweave.net/afjNOdumqZhfVMIjRw_8fCYZoecSB6XtHGZpNNUNIr8/mod.js

With nothing else, if we go directly to Arweave using g8way.io, we lose visibility into the module_name and version. We need a user friendly domain that identifies the module name and version and then use that to import from in Deno. For that, we could use https://ar.io/arns

We already own the hyper Subdomain on hyper.g8way.io, so by "registering" subdomains like adapter-sqlite-1.2.3_hyper that point to the transaction that contains the source code, we could have user-friendly urls to import from like import cache from https://adapter-sqlite-1.2.3_hyper.g8way.io/mod.js

This is still all vanilla Arweave and SmartContracts

[TillaTheHun0]

So to deploy CI could run a script, ran on new git tags, that:

• run bundlr.uploadFolder being sure to omit any secrets files that were injected into CI (currently none)
  • maybe could set mod.js as the indexFile
• run setRecord for hyper ANT contract using the kebab-case of the package name+semver ie. adapter-sqlite-1.2.3, pointing to the transaction id of the manifest file generated by bundlr.uploadFolder.

A separate manual task once this is done

• Update README with new import url schema ie. import cache from https://adapter-sqlite-{semver}_hyper.g8way.io/mod.js