Agent vs Client is a bit confusing
tigrannajaryan opened this issue · 3 comments
The spec currently uses the term Agent mostly (although Client can be seen in the Communication Model). The Go implementation on the other hand uses the term Client.
This is a bit confusing.
Should we mostly use the term OpAMP Client and only where relevant talk about the Agent?
This is especially more important now that Supervisor seems to be a likely option to implement OpAMP for Otel Collector. It makes things even more confusing. Unclear whether OpAMP Agent is (Supervisor+Collector) or just Supervisor.
In the 8/9 agent management meeting we agreed that Client is the preferred term.
One of the arguments for "Client" is that it well describes the role this endpoint plays in the protocol, while "Agent" suggests functionality, which, while important, is not really defined by the protocol. Furthermore, Agent might become even more confusing if we will have Tracer/SDK implementation(s) of the client side of the protocol in the future.
@andykellr I agree. We should probably create a PR that tries to rename Agent->Client where relevant (while still keeping Agent where it actually refers to the Agent). We should probably also rename the Protobuf messages (e.g. AgentToServer becomes ClientToServer, etc).
Do you think you will be able to submit a PR like that? If not then I can try to do it myself in a couple weeks (busy with something else right now).
We should do a draft PR first to see what it looks like. I am not entirely sure it will work nicely, but we should try.