On 5/3/11 11:20 AM, Anders wrote: > On 04/28/11 18:29, Linus Lüssing wrote: >> In my opinion, that should be okay, but I'm also using IPv6 >> regularly. What would be the alternative, installing the bridge as >> a kernel module instead of kernel built-in as it is currently done? >> >> Hmm, if no one is screaming about any space issues and not wanting >> to have IPv6 installed, I guess it should be fine to have the IPv6 >> code built-in, shouldn't it? >> >> > Screaming? No, but it would be good to see what the cost in space was > before comitting it. > > The old 2.4 kernel left ~1.5M fee on my linksys, the 2.6 one redices it > to 800K, so space is, > or will soon become, an issue, and it's a still a (largely) IPv4 only > world out there. > > -A
Injecting myself into this conversation: "it's still a largely IPv4 only world out there" will continue to be the case as long as user devices are not ubiquitously IPv6 capable: that's a truism. I've had to deal with similar circular reasoning when trying to get various packages out there to support DSCP packet marking: the claim being from application vendors (including the OSS community) that "networks don't support DSCP/QoS". In my conversations with ISPs and IXCs, I hear the rational, "we don't deploy QoS support because not enough applications leverage it to make it worth the (not insubstantial) incremental cost". Since the software cost on the user side is a lot cheaper (and more scalable), let's break the deadlock/chicken-and-egg-dilemma there... -Philip _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/mailman/listinfo/openwrt-devel
