Nice useful timeouts

[ghost]

Note: I'm only talking about /ipfs and /ipns paths on the gateway here, not about /api.

Currently there are hardcoded timeouts of 1 hour in the gateway, and 30 minutes in the nginx reverse-proxy on ipfs.io. If the gateway's underlying ipfs node can't serve the file/directory from its own dagstore, it'll use bitswap to find and fetch it. The request will hang and wait until the data has been fetched. If one of the timeouts is reached, the response will be a cryptic error ("context deadline exceeded" and "504 Gateway Timeout", resp.)

We can make this condition a lot more useful.

• Low request timeout in the gateway, something like 30 seconds
• A timeout response page with:
  • "This is what might be wrong"
  • UI for retrying, and setting ?timeout query parameter, one-shot or permanently
  • Automatic polling for the requested object (via /api), with progress feedback, offer to reload when finished
  • UI for contextual debugging, e.g. dht findprovs/findpeer, name resolve, wantlist, etc. (via /api)

Basically like a search form on a 404 page, but way cooler.

[ghost]

To further improve NAT traversal debugging, we could probably have the 504 page exploit CSRF vulnerabilities in plenty of home routers to check the model against a list of known-bad devices (regarding NAT traversal). Just kidding :)