adlrocha
commented
3 years ago Just wanted to drop by to let you know that I love where this is going. I wish I had had more time to approach this refactor. One of the initial goals for the testbed was to have the environment for anyone to be able to come with her or his exchange protocol implementation for IPFS (or their improved Bitswap/Graphsync), plug it into the testbed, and evaluate their performance against existing protocols. This refactor is definitely a step towards that vision.
A further refactor is to make a common executor, that maybe takes a fetch function if we don't want to put it in the node. Perhaps it all becomes a single test case except for TCP? It certainly would make the testplan easier to read.
And I think this would be something awesome to have. It would make the testplan easier to read and the testbed way more flexible. Let me know if you need additional input or a quick review from my side.
Goals
Unify the node interface for bitswap & IPFS transfers, approach unification of testcases for IPFS/graphsync/bitswap
My hope here -- and I haven't tackled whether we want the nodes to all have a Fetch method -- is that we can get to the point where if you can satisfy the Node interface, you can run it against the test infrastructure
Implementation
For Discussion
transfer.go, ipfsTransfer.go, and graphsyncTransfer.go now look nearly identical. The only major difference is the fetching code. Which could be put in the node likely (thought there are interesting challenges cause some need a targer peer while others don't)
A further refactor is to make a common executor, that maybe takes a fetch function if we don't want to put it in the node. Perhaps it all becomes a single test case except for TCP? It certainly would make the testplan easier to read.