Closed Evanfeenstra closed 12 months ago
@Evanfeenstra upon further thought, it seems to me we still need that BUSY atomicbool to make sure that handle_message
is done by only one thread at a time right ?
We want a spinlock and a global counter.
@Evanfeenstra upon further thought, it seems to me we still need that BUSY atomicbool to make sure that
handle_message
is done by only one thread at a time right ?We want a spinlock and a global counter.
i guess so, makes sense
from devrandom
the [LSS] state is only used for two things:
- when first building the Handler / Node / channels
- when deciding what changed when running an API call
- Right now for a multi-signer setup you would need to rebuild the root handler completely
as a reminder, we didn't design for multi-signer (yet)
On the second point - is that the signer computing the "state diff" to send back to the LSS server?
On the second point - is that the signer computing the "state diff" to send back to the LSS server?
exactly
Section 1 addressed in #95
use the broker/looper
sequence
to verify on the signer side that no VLS message was skipped. Each message should increment by one.sequence
from the individual Channel structs.sequence
is 0 OR jumps more than 1, then the signer needs to do the reconnect dance to sync the state, and then finally process the VLS messageBroker should send the VlsMuts(SignerMutations) to all other connected signers whenever it receives it from one