Re: uptime (coreutils)

[Thomas Bushnell, BSG]

Roland McGrath <roland@frob.com> writes:

> libc
> libm
> libdl
> libutil
> libresolv
> libcrypt
> libncurses
> libreadline

Oh, of course; those aren't system-specific libraries though.  It's
the idea of splitting libc into different system specific libraries
that I think is a mistake.

I have no objection to a new generically named library for specific
sets of tasks.

(Of course five of the libraries in your list are all generated from
the libc source. :)

> Well, you used to be named Michael; I think my vacillation is minor in
> comparison.  

I don't object to changing of minds; that's fine.  

> I think there ought to be a separate Linux library for purely
> Linux-specific interfaces that other system could possibly implement.  It
> just makes life simpler for writing portable applications.  There are only
> a few pieces of a GNU/Linux ABI that are hard to match exactly on a
> non-Linux implementation.  It just makes sense to segregate those so their
> library (or at least version set) dependencies indicate clearly what kind
> of nonportable hooey they rely on.

Except I think it's a mistake to label those "linux-specific".  The
danger of that is that as soon as it gets implemented on some other
system, it ceases to be Linux specific.

Moreover, there is nothing Hurd-specific about fetch_boot_time.  It
should be implementable on any system, and is a reasonable entirely
generic function.

Thomas