On Tue, 2019-12-17 at 21:11 +0000, Stotland, Inga wrote: > Hi Michal, Brian, > > On Tue, 2019-12-17 at 21:06 +0100, michal.lowas-rzechonek@xxxxxxxxxxx > wrote: > > On 12/17, Gix, Brian wrote: > > > > Indeed. Would you be willing to accept a patch that changes token > > > > delivery to use JoinComplete call in all cases, and checks that the call > > > > returns successfully? > > > > > > Maybe... I need to discuss it with Inga. How does this relate to > > > Create() though? There is no JoinComplete() call currently axs a > > > result of Create()... just the return value from the Create() method. > > > Is your problem that the daemon cannot be sure that the return value > > > (token) from Create() was received? > > > > Yes, exactly. > > > > As much as I don't like API changes in a released code, this seems to > be a legitimate case for doing so. > > So yes, let's change the logic so that all three methods (Join, Create, > and Import) have void return values and the token is delivered in > JoinComplete(). A node is "finalized" only upon successful return of > JoinComplete(). And if the call to JoinComplete() fails, the daemon should delete the node storage. > Best regards, > Inga
