Provide direct support for Best-Effort pins?

[kevina]

Should we provide direct support for creating best-effort pins? In some ways it might not be necessary because users can just add stuff to the MFS root that they want to keep around but don't want a recursive pin for.

Related to #3069.

[whyrusleeping]

Yeah, i'm in support of having best effort pins. Although i think that in order to do that, we might want to first rethink how pins are stored. Take a look at #1580. That bit might be a really nice addition to have before best effort pins because then users would be able to ipfs pin add --type=best /ipfs/<hash>/path/to/dataset and be able to still resolve that full path after a gc run.

[lidel]

One can do best-effort pinning like this:

1. ipfs files cp /ipfs/{cid} /mfs-destination will add DAG to MFS, acting as "implicit" pin that protects from GC
2. ipfs pin -r {cid} (or ipfs dag stat {cid} – preload blocks without explicitly pinning, use MFS for GC protection)

My understanding is that if you interrupt (2), the (1) will protect blocks that are already fetched.

I am not sure how valuable is adding best-effort mode to ipfs pin unless it is the default behavior. When one needs to do some sort of opt-in, MFS (ipfs files cp) feels more intuitive to me, but that could be subjective. Thoughts?