Bernd Eckenfels wrote: > In article <000901c2aabb$eb2d0e40$210d640a@unfix.org> you wrote: > >> no they didnt, I must know I am the net tools maintainer. The > >> default target is supported in route 1.57. > > > > Hmmm you probably know this better but as (cut from route.c) > > 8<--------- > > * (derived from FvK's 'route.c 1.70 01/04/94') > > ---------->8 > > I think one can say that it supported it all along :) > > 1994 is almost 10 years ago again, and the patch went in before > > that apparently... > > the CVS versions of the source files are not related to the net-tools > release version. 1.57 was released in May 2000. > <http://www.tazenda.demon.co.uk/phil/net-tools/> That explains, and I prolly didn't notice at the debian packages (provided by you to debian ;) get upgraded automatically without me checking what gets upgraded. btw there is no seperate changelog... > >> it is not a hack it was there for a good reason, to avoid > >> link and site local routes. > > > > No the 2000::/3 is a hack because one couldn't specify a default > > route using the <2.4.20 kernels. > > the reason why the default route for gateways(!!) was not > possible is not a bug or missing feature, but by intention, > to avoid a route on a exterior gateway which would also > route site and link local addresses. That explains the behaviour a bit, even though I prefer the second version, ability to set default routes and then fixing the site+link locals seperatly usually per interface. > > It has, as far as I know and realize, always been like this > on Linux: > > yes because site and link local routes are not matched by 2000/3 > > 2000/3=0010xxxx... > fe80 =11111110.... This could break one day (far far away when our childrens kids grow up :) Btw it's 001xxxxxxxx (0010 == /4). In the far future one could see the 010/011 etc variations also being used if the address allocations seem to go wrong etc. But it looks like the current allocationscheme is doing quite well. > > As you see I still have a 2000::/3 because a default can't be added > > because of the <2.4.20 kernel and the fact that this box > has forwarding > > turned on. > > yes, but if you add 2000::/3 you do not have a default route > which routes site and link local addresses, which is the idea > behind using that prefix. > So it is not a good idea to actually use "default" aka 0::/0. > (unless 2.4.20 is not routing those prefixes, which I have to check > in the code -or- you manually make sure you habe site/link local > point to somewhere else). Usually, as far as I know on any linux kernel sporting v6 one will see something similar to: 8<----------- 3ffe:4007:1:1::/64 dev eth0 proto kernel metric 256 expires 2591946sec mtu 1500 advmss 1440 fe80::/10 dev eth0 proto kernel metric 256 mtu 1500 advmss 1440 ff00::/8 dev eth0 proto kernel metric 256 mtu 1500 advmss 1440 default via fe80::230:7bff:fe19:f400 dev eth0 proto kernel metric 1024 expires 1746sec mtu 1500 advmss 1440 unreachable default dev lo metric -1 error -101 ------------>8 So the fe80::/10 and ff00::/8 prefixes are routed over eth0 and not to the 'outside'. This holds true for (afaik ;) gateways and clients. Also afaik this has always been this way. > > Anyone know the real reason ? > > trust me :) I don't have any problems with that ;) And it's good to know the real story about the non-default-on-gateways. Greets, Jeroen - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
