> In article <Pine.LNX.3.95.971031112339.10824A-100000@ferret.lmh.ox.ac.uk>,
> Chris Evans <chris@ferret.lmh.ox.ac.uk> wrote:
> >Features that seem to be essentially "done" code-wise, and I would like to
> >see in the kernel include
> >
> >1) The new RAID levels
> >2) ACL's in ext2
> >3) New filesystems -- HFS, NTFS
> >4) The "privs" root-splitting part of posix.1e
>
> Some more random thoughts:
>
> o Online mirror daemon (the reason for the dentry overhaul)
> o Merging ppp and syncppp, creating modular interface
> o SCSI interface overhaul for error handling etc
> o ext2: compression, ACL
> o POSIX signals
> o If POSIX signals not implemented, alternative way for syncing
> pthreads (linuxthreads) instead of using SIGUSR1 and SIGUSR2
> o NAT
2.1 has already simple NAT - see net/ipv4/ip_nat_dumb.c. There is
just no CONFIG_ option for it, it's always compiled in so people
never noticed it. In DaveM's CVS tree there is a CONFIG_ option,
so people will start notice it. The NAT is called dumb BTW, because
it doesn't support the protocol modules used by the masquerading code.
That means that protocols like FTP or IRC that pass IP addresses don't
work over NAT yet. I think it would be fairly easy to add a interface
to the masquerading modules to the ip_dumb_nat code (and make it a
ip_smart_nat :). Any takers?
Oh, yes, strictly speaking masquerading is NAT too - just a special
case of 1:N NAT.
Regarding 2.2: My personal hope was that 2.2 would have more code outside
the giant lock for more parallelism in the SMP case. But it seems it's
too late now :(
-Andi