Issue when setting delay=0 to NetCon and multiple cores

[danilobenozzo]

https://github.com/dbbs-lab/bsb-neuron/blob/fc35e97006692c2e3aa1a570ce49b01c28d29b6d/bsb_neuron/connection.py#L30C5-L30C49
I think this overwrites the default values that were initialized here https://github.com/dbbs-lab/patch/blob/4d7442f726a556d28a7b72629baf3d427fac6568/patch/interpreter.py#L134 placing the NetCon.
A delay by default equals to 0 will give this NEURON error: usable mindelay is 0 (or less than dt for fixed step method) when running in parallel.
Probably better to keep the same default values as in patch/interpreter.py ?

[Helveg]

Yes, a higher default would be better. It may even be best to make it a required argument without a default at all? Or default value None so that it indeed uses the nrn-patch default values.

[danilobenozzo]

In the case of a transmitter sending to multiple receivers delay and weight cannot be receiver specific because they refer to the NetCon which is in the transmitter, right? So making them required arguments might be misleading for the user.
I prefer default value None so that the nrn-patch values are used, but should a is None check be added in patch.interpreter.ParallelCon()? As it is now, I think it will assign None

[Helveg]

In the case of a transmitter sending to multiple receivers delay and weight cannot be receiver specific because they refer to the NetCon which is in the transmitter, right?

Hmm, indeed ... but that would mean that if you have a connection from section A to B and section A to C that they can not have different delays. That seems like an impossible restriction for NEURON, especially when you know that in NEURON 99.9% of users have strictly 1 NetCon on the soma of each cell. There must/might be a workaround. Maybe:

• We can register a delay on the receiving side, somewhere in the NEURON calls that ParallelCon makes?
• We can register 2 different GIDs to the same NetCon, with different delays?

Otherwise, ask the NEURON devs how to deal with the situation where 2 different delayed spike signals need to be emitted from the same Section.

---

Once you know more, it will become clear what the best solution is :)