----- Original Message ----- From: Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> Date: Friday, March 28, 2008 12:51 Subject: [haiku-development] Re: GSoC ICMP implementation To: haiku-development@xxxxxxxxxxxxx > "Alvarez, Alan SPC MIL USA FORSCOM" <alan.alvarez@xxxxxxxxxxx> wrote: > > The problem with the approach you originally suggested is that > > sometimes the error is > > caught at the Session layer (TCP/UDP) and by then the header has > > already been removed. > > But is it necessary to include the original header in these cases, > too?If that is the case, I think your idea of having it clone the > header > first would work. It's just the question on how the ICMP module > can > then access it, and when that header can be thrown away; it seems > like > this needs some changes in the current infrastructure. > Eventually, one might want to attach additional data to a net_buffer. > yes. The IP header has to be sent back the exact way it came. Well, it could be required for an upper module to call a lowest module to "clean up". I'm not sure if this is a very clean design though. I'll think in the next few days and let you know what I can come up with. > > I'm not sure how I could call the icmp error handler separately > and > > before the header is > > cut off. I'm sure I'm missing something. > > You mean in the case above? That wouldn't be possible, I guess :-) I thought I was going crazy for a while. > > > I'm sorry for dragging this for so long. If you wish we could > resume > > this if/when the project > > gets accepted. > > I think it's always better to sort out things like this as early > as > possible. I don't have a deep knowledge about ICMP anyway, so I > can > only benefit from yours :-)) > > Bye, > Axel. > > > - Alan Alvarez