Storing Neuron Ids

[dfx-json]

From @bitdivine

Hello. Just an idea. I have been in the habit of storing neuron IDs in ~/.config/dfx/identity//neurons/<network_name> . I do this as I often need to know the ID of a neuron I control. Other people might potentially find something similar useful. It might be worth having a short design session and aligning on a strategy rather than having lots of individual solutions though. E.g. how do we want to nuke neurons and wallets when a testnet is torn down? Do users want a default neuron? (I do) And so on.

[dfx-json]

Response from @ericswanson-dfinity

We sometimes invite someone from outside the SDK team to present at an architecture meeting for things like this. Would you like to do so? The agenda is here: https://docs.google.com/document/d/1M3LXK9mM19Yhz6wZPJCh7V9zvIj_GokkV5hyq2lIaIk/edit?usp=sharing . A good place to start before such a presentation might be to write up a 1-pager.
The first question I would like to have answered is: Is dfx (and its data files / configuration files) a good fit for this? Context: Quill manages neurons, while dfx does not. It even turns out that the only references to neurons in the sdk repo are in the dfx sns and dfx nns code, which we are in the process of moving elsewhere. Followups: Do we foresee adding neuron-related commands to dfx, or only to quill? Would dfx then be managing these neuron ids for use with quill?
I think the connection to wallets is insightful: That's another place where we associate (identity, network) -> identifier, and which we need to clear in the case of dfx start --clean.
It may be useful to consider the characteristics of the wallets.json file here:
Locations:
~/.config/dfx/identity//wallets.json (persistent networks like ic and testnets)
~/.local/share/dfx/network/local/wallets.json (local shared network [linux])
~/Library/Application Support/org.dfinity.dfx/network/local/wallets.json (local shared network [mac])
/.dfx/local/wallets.json (project networks)
File format is the same in any of the above locations, so that after determining the file location, operations on it are the same.
Cleanup operation is deletion of a single file rather than of a directory and its contents.
Internally, the file structure is identity name -> network -> identifier
I would encourage the use of a single file following the same model, unless there is a good reason for some other scheme.
(and a note: while it may be unlikely/uncommon to store neuron ids for a local network, for consistency dfx would still have to handle that case)
As you mention, dfx does not provide a way to clear wallet data when tearing down a testnet (which is a "persistent" network, yet sometimes torn down), and nor would it for neuron ids. To me this sounds like two proposals that would be best separated: 1) Can(/should) we store neuron ids with dfx data/configuration, 2) Some new dfx command to clear relevant data when tearing down a testnet.