I hope I'm not repeating something said elsewhere in this rather
large thread. I just wanted to point out a distinction that seems
to have been missed so far.

Dynamic linking may well be required for dlopen().

A number of examples have been given (perhaps in more detail in
another thread some time back) of why having dlopen() is desirable;
the most prominent being locale support.

However, dlopen() is not strictly required for locale support, even
with its full intricacies beyond simple message catalogs as
discussed here.  It's only necessary for *dynamic* locale support.

Those of us with English-speaking backgrounds have been quite
successfully using these static tools with static localisations
for some time.  It's no surprise that the folks wanting to maintain
the status quo fall into this group. They're never going to want
to use /sbin/restore in Korean.

I just wonder whether someone in Korea, in the middle of trying to
recover a messed-up system, would ever suddenly decide to switch
to Polish?

I can hear releng screaming already, but allowing binaries to be
statically linked (or patched at install time?) with a  chosen
locale doesn't seem far-fetched - even if they live in /rescue on
an otherwise fully dynamic system.

Where does that leave other potential uses of dlopen() for the base
system tools?

--
Dan.