Oliver Tappe <openbeos@xxxxxxxxxxxxxxx> wrote: > On 2006-07-31 at 16:32:18 [+0200], Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> > > wrote: > > If we do that, we can for sure get rid of address_offset and > > address_length; they are only used for comparison and (not yet) > > masking. It probably depends a bit on the needs of other address > > domains if this is really necessary (in terms of required), anyway, > > but > > I must admit I haven't had a closer look at IPv6 yet (I just went > > the > > BONE route :-)). > I have just read up on IP6-addresses and while the subnet masking > seems to be > the same (CIDRized network addresses, only 96 bits longer >;o), but > there > seem to be subtle differences in the differentiation between > broadcast, > multicast and unicast addresses. But we can always move that code out > into a > special address-module when we finally tackle other domains and > consider it > worth it. So we'll keep my mess for now? Damn, I thought you would clean it all up... ;-) > What worries me more is that I didn't pay any attention to keep my > UDP > implementation address-type agnostic. There's sockaddr_in all over > the place > and that would mean that we'd have to reimplement UDP for the IP6 > domain. > Bummer! >:o/ > I'll better clean this up now, as it will only hurt more to do it > later... You probably better know how IPv6 affects UDP. But since you do IP "arithmetic", too (at least you're comparing addresses :-)), maybe it would be nice to export such functions from the datalink module, or go the full monty and have that modular as you suggested in the first place. Bye, Axel.