More strict MV3 rules for path and subdomain gateways

[lidel]

URL like https://ipfs.io/ipfs/QmbWqxBEKC3P8tqsKc98xmWNzrzDtRLMiMPL8wBuTGsMnR produces rules like

origin: '^https?\\:\\/\\/ipfs\\.io',
destination: 'http://127.0.0.1:8080'

The MV3 rules created by ipfs-companion beta 3.0.0.1114 for a public path or subdomain gateway should be for specific namespace like /ipfs or /ipns, never for the entire origin. Redirecting entire origin should happen only for DNSLink websites.

For public path and subdomain gateways we don't want to have a situation where we break delegated routing because companion hijacked /routing/v1 endpoint, or some other /foo service endpoint that domain owner decided to expose on the same hostname,

Tasks

Beta Give feedback

1. Make rules less greedy and explict to the namespace.
2. Add unit tests for DEFAULT_LOCAL_RULES feat(mv3): Redirection Tests #1236 (comment)
3. Don't break dev environments: feat(mv3): Redirection Tests #1236 (comment)
4. MV3 handling of ?uri= creates too many rules #1254
   mv3-beta-bugs +1
   
5. MV3 handling of subdomain gateways creates too many rules #1255
   mv3-beta-bugs +1
   

[whizzzkid]

@lidel came across this fleek example which does not have a namespace in the url.

• https://helia-identify.on.fleek.co/ -> http://localhost:8080/ipns/helia-identify.on.fleek.co/
• https://bafybeib3bzis4mejzsnzsb65od3rnv5ffit7vsllratddjkgfgq4wiamqu.on.fleek.co/ -> http://localhost:8080/ipfs/bafybeib3bzis4mejzsnzsb65od3rnv5ffit7vsllratddjkgfgq4wiamqu/

The regex pattern that will be generated for 2nd https://(.*?).on.fleek.co/(.*?) -> https://localhost:8081/ipfs/$1/$2 interferes with the first, so that rule cannot be dynamically generated 🥺

[lidel]

My suggestion would be to

• remove special-handling of Fleek (even if it is "generic" magical detection of CIDs in domains that do not have namespace label)
• reach out to Fleek and ask them to follow the specs (linked below) and use {cid}.ipfs.on.fleek.co instead

We have subdomain convention for a reason, it requires namespace label after the CID to remove surface for false-positives (confusing with ipns, peerid, or uses outside content-addressed paths):

• https://specs.ipfs.tech/http-gateways/subdomain-gateway/
• https://docs.ipfs.tech/how-to/address-ipfs-on-web/#subdomain-gateway