On Sun, 4 Mar 2001, Arkadiusz Miskiewicz wrote: > On/Dnia Sun, Mar 04, 2001 at 03:45:27PM -0700, Jason Gunthorpe > wrote/napisa�(a) > > The trick is that if you request a passive socket from getaddrinfo, it > > should return a list of sockets that should *all* be bound.
> Why you think that it *should*? As I said there is no standard about this. > Also why it should do this only when AI_PASSIVE is passed? Why not? That seems to lead to the nicest implementation, and makes alot of sense. > IMO only AI_ADDRCONFIG flag is better resolution (AI_ADDRCONFIG|AI_PASSIVE). > We get more flexible getaddrinfo(). But that doesn't actually fix the problem. AI_ADDRCONFIG|AI_PASSIVE still has to return either v6, v4 or both depending on what kernel it is for proper operation. Arguably though AI_ADDRCONFIG|AI_PASSIVE should have the behavior I am describing.. Jason
