Go-ipfs: `ipfs p2p forward` should accept full multiaddresses as target (not only /ipfs/<peerID>)

[hsanjuan]

Version information:

master

Type:

Feature

Description:

ipfs p2p forward takes an */ipfs/<peerID> multi-address as target. This assumes the given peer is routable.

However, sometimes it would be useful to do ipfs p2p forward <proto> /dnsaddr/mypeer or ipfs p2p forward <proto> /ip4/1.2.4.5/tcp/23131/ipfs/<peerID>. In these case, the command should automatically add the peer addresses to the peerstore so that they can be connected (right now, a manual swarm connect is required in advance). resolve the addresses.

This simplifies connecting to libp2p peers which are not part of the ipfs network or, particularly, the ipfs dht.

[magik6k]

If no one grabs this soon, I'll do that in a spare moment

[kjzz]

Hey @magik6k @Stebalien , can i help you do for this?I am interested in this issue.

sometimes it would be useful to do ipfs p2p forward /dnsaddr/mypeer or ipfs p2p forward
/ip4/1.2.4.5/tcp/23131/ipfs/

Can you explain it to me?Thx a lot.I am confusing about this.

And as i known, we can use command ipfs p2p forward /x/_testing /ip4/127.0.0.1/tcp/4567 /ip4/127.0.0.1/tcp/23131/ipfs/<peer id> and ipfs p2p forward /x/_testing /ip4/127.0.0.1/tcp/4567 /dnsaddr/bootstrap.libp2p.io/ipfs/<peer id> to connect right now.

[hsanjuan]

@kjzz you're right, it already takes a full multiaddress and adds it to the peerstore before forwarding so that should work. What doesn't work is resolving /dns*/ first. Sorry for the confusion here.

[kjzz]

Thx for replying @hsanjuan , so you mean that we should support command such as ipfs p2p forward /x/_testing /ip4/127.0.0.1/tcp/4567 /dnsaddr/bootstrap.libp2p.io?
And before forwarding , we should use resolve function to get all multiaddrs?Then add them all to peer store?

[hsanjuan]

@kjzz yes, in principle, but we have some discussion about how painful it is to manually resolve in several places etc. https://github.com/ipfs/go-ipfs/pull/5535#issuecomment-425597268 maybe hold off a bit, as ideally this should be taken care of directly at libp2p-level.