Re: Path to kmods

[der Mouse <mouse%Rodents-Montreal.ORG@localhost>]

>>>> I fully expect non-dynamically-linked kernels to be desupported
>>>> within some small number of years;
>>> That is unlikely to happen given the market segments NetBSD does
>>> well in.
>> I hope you're right.  But I look at what's happened to userland
> what, exactly, has happened to userland?

More and more things being linked dynamically.  Back in the day, there
was no dynamic linking.  Then we got dynamic linking, with stuff in
/bin and /sbin still linked static.  Then the static binaries were
pushed into /rescue, so /bin and /sbin could be linked dynamically (I
still don't see why this is better than leaving static binaries in
/{,s}bin).

Now, see tech-toolchain from mid-August: someone proposed dropping .a
libraries from NetBSD-shipped binary distribution.  (It got shot down,
but that it got taken seriously enough to need shooting down is what I
find significant.  In case you want to find it on mail-archive, the
Subject: header I see in my mail history is "Stop shipping static
libraries for NetBSD".)

I see a clear direction here - and I see it spreading to the kernel.

It started with kernel modules, the first introduction of dynamic
linking to the kernel.  Now I'm seeing it following more or less the
same progression.  I expect NetBSD to ship systems that no longer
support userland dynamic linking within a year or two.  Systems that no
longer include statically-linked anything within six months to a year
after that.  Another six months and static-linking support will have
bitrotted into uselessness (since it's no longer widely used) and there
will be calls to remove it entirely.  The kernel is just lagging a bit
behind.

I hope I'm wrong.  But that doesn't mean I'm not thinking about what to
do if it turns out I'm right.

/~\ The ASCII                             Mouse
\ / Ribbon Campaign
 X  Against HTML                mouse%rodents-montreal.org@localhost
/ \ Email!           7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B