[Tfug] linux no longer for amateurs

Bexley Hall bexley401 at yahoo.com
Sat Jan 16 09:26:53 MST 2010


Hi Zack,

> Bexley Hall <bexley401 at yahoo.com> wrote:
> > Zack said:
> >
> > Since then, I haven't deviated from distro releases. There's a reason
> > that we have them, and not just a kernel with separate userland - none
> > of the BSD's or any other commerical unixes do it the linux way.
> >
> > Frankly, I didn't know if that was praise for Linux or derision.
> > I erred on the side of assuming there was a reason *why* Linux
> > does things "the linux way".  I asked if he would care to
> > explain *why* this was "better" (though he could just as easily
> > have explained why it was *worse*!).
> 
> Ah, damn, I should have been more clear - I intended that
> as a praise for disro releases, if only because they get more Q&A
> and another set of eyes over everything when it's integrated.   

Understood.  It's *a* configuration that AT LEAST ONE PERSON
has (presumably) worked with and assured him/herself that it
is functional.  (Or, has a finite set of bugs  :> )

> Thus, they're more likely to work, in my experience.
> 
> To use a car analogy, the Linux method is to sell a crate engine, and
> have other people build the rest of the car, whereas everyone else out
> there builds the whole car, or uses a specific crate engine that
> they're sure works correctly for them.

I'm not sure the analogy holds.

E.g., the *BSDs are "complete" OS's -- not just kernels
(Linux, IIRC, is *just* the kernel).  But, there is also
a distinction made as to what is part of the "OS" and
what is "application".

So, a *BSD will contain some number of prebuilt /*bsd kernels
(e.g., netbsd.GENERiC, netbsd.INSTALL, netbsd.TINY, etc.)
along with gcc, inetd, init, vi, fsck, hangman, etc.

But, it won't contain kde, asterik, OO, tetris, etc.

From what I've gleaned ("overheard"), it seems like a distro
is "all of the above" -- kernel, "the rest of the OS" *and*
applications.  Though the set of applications offered will
vary (?) from distro to distro (i.e., one might use desktop
A while another uses desktop B, etc.).

But, I don't see how any of that pertains to problems
building the kernel, itself.  Or, why a successfully built
kernel wouldn't work with <pick_one> distro?  I can see
how a kernel that doesn't have a sound device in it
would render all sound apps inoperable.  This is true
in *BSD-land as well -- and is intuitive.  E.g., *BSD
requires certain kernel support for X -- omit that
support and all of the X-related stuff is dead.

But, the OP (eric?) seemed to suggest that building *a*
kernel for "reasonable" hardware was a serious chore...?


      




More information about the tfug mailing list