The initial implementation of IPNI used GossipSub as a way to propagate
IPNI advertisement chain. To do this the propagation had to be relayed
through the Lotus node due to strict Filecoin GossipSub validation
rules.
Since then IPNI has moved on to roll out its own sync protocol that
works over HTTP, and HTTP-over-libp2p. This move has been the official
way of advertising content to IPNI federation over a year now.
Therefore, remove the ad relay over pubsub via Lotus node as it is now
considered to have reached its EOL as a mechanism for advertising to
IPNI.
This is a large diff, yet should have exactly zero functional changes
Ideally as a result of this some parts of the depchain will become lighter,
with downstream reaping the same benefits as the team that initiated this split.
P.S. work was done while forming better intuition of current dependency graph
* remove client CLI
* remove markets CLI from miner
* remove markets from all CLI
* remove client API
* update go mod
* remove EnableMarkets flag
* remove market subsystem
* remove dagstore
* remove index provider
* remove graphsync and data-transfer
* remove markets
* go mod tidy
* fix cbor gen deps
* remove deal making from config
* remove eol alert
* go mod tidy
* changes as per review
* make jen
* changes as per review
* merge master
* remove libp2p from config
* miner does not have libp2p conn in api test
This reverts commit 8b2208fd9a3d70fa4ef419c12a3953337d45c807, reversing
changes made to 2db6b12b78baf6e73cfa16a86ae7e267fd967421.
Unfortunately, this is rather tricky code. We've found several issues so
far and, while we've fixed a few, there are outstanding issues that
would require complex fixes we don't have time to tackle right now.
Luckily, this code isn't actually needed by the main Filecoin chain
which relies on consensus fault reporting to handle equivocation. So we
can just try again later.
Content providers announce the availability of indexer data using gossip pubsub. The content providers are not connected directly to indexers, so the pubsub messages are relayed to indexers via chain nodes. This PR makes chain nodes relay gossip pubsub messages, on the /indexer/ingest/<netname> topic.
LifecycleCtx can _only_ be called during startup as it appends an fx hook.
Worse, this was causing us to append an fx hook on every single hello message,
leaking memory (and probably causing other shutdown issues...).
We were ignoring quite a few error cases, and had one case where we weren't
actually updating state where we wanted to. Unfortunately, if the linter doesn't
pass, nobody has any reason to actually check lint failures in CI.
There are three remaining XXXs marked in the code for lint.
