svn commit: r243594 - head/sys/netinet

[Adrian Chadd]

..right. I could spin the argument back to you Alfred and say "make
the default compile time settings for i386/amd64/sparc64 GENERIC to
have bigger TCP hash sizes, default mbuf tuning, larger kernel dmesg
buffer sizes, increased mutex hash table sizes, etc, and leave the
defaults as they are.

What I'm trying to say/encourage is some more even-handed ways of
doing this, rather than "hey, it works in the server world, that's all
that matters, right?"

And FYI - installation on embedded devices these days is "open device
web page, upload firmware image, hit "flash", reboot." That's what
openwrt and dd-wrt do for the majority of their platforms. No serial
console required.

As I said, I don't think everyone invested in this is seeing the
bigger picture and trying to work out whether we can solve this in a
better fashion. The peeps in the Linux world _have_ come up with _a_
solution that seems to work. So it's not that its intractable.


Adrian