XPost: comp.os.linux.development.system   
   From: nate@vulcan.lan   
      
   phil-news-nospam@ipal.net writes:   
      
   > In comp.os.linux.development.system Nate Eldredge  wrote:   
   >   
   > | I can tell you what FreeBSD does.  The kernel config file specifies   
   > | various options and lists the drivers that are to be compiled in to the   
   > | kernel.  Every driver not listed is compiled as a module (provided it's   
   > | appropriate for the architecture, e.g. the Sparc firmware drivers are   
   > | not built on i386).  So no matter what, all reasonable drivers are   
   > | compiled and can be used, whether built in or as modules.   
   > |   
   > | Changing the set of devices that's to be compiled into the kernel   
   > | requires manually editing the config file; it isn't done automatically.   
   > |   
   > | I honestly can't understand why you think it is practical or desirable   
   > | to automatically create a kernel with the "right" set of drivers   
   > | compiled in.  The whole point of the module system is that this is hard   
   > | to do at compile time, and is better left to run time.   
   >   
   > Install time *IS* a run time of the installer system.  So it can be seen at   
   > that time what the target machine needs.  If for some reason that is a   
   problem   
   > then perhaps some distribution might defer the kernel rebuild until after the   
   > first boot.  Then it can see what modules got loaded (e.g. what was "right"   
   > at run time), and configure a kernel that differs from the currently running   
   > one by making just these drivers static.   
      
   My point was that the set of necessary drivers can change from run to   
   run, and so it's not sufficient just to use the set which appears to be   
   needed at the time the installer is run.   
      
   > All I asked for was if any distribution existed that did this.  It should   
   have   
   > been a simple answer.  If you knew of no such distribution, you could offer   
   > that thought if you were confident your knowledge of distribution was wide   
   > enough that you would likely have seen it if it did exist.   
      
   I don't know of any such distribution.  I am not so confident that I   
   would know if one did exist, but I would be surprised if one did.  As I   
   said, I don't see how this behavior would be useful, and therefore I   
   don't see why any distribution should implement it.   
      
   > I did not ask this   
   > question in order to get a bunch of answers to other questions people   
   presumed   
   > that I should have asked, based on the (false) assumption I have never built   
   a   
   > kernel and do not know how.   
      
   That wasn't my assumption, by any means.   
      
   I couldn't understand why you wanted what you asked for, and so I probed   
   a little more.  Partly I was curious if there was some reason for it   
   that I hadn't thought of, and partly I wanted to know what your   
   underlying goal was, in case I knew some other approach to it.   
      
   > What does it take in Usenet to get people to either answer the question that   
   > was asked or to now answer at all, but in particular to not make presumptions   
   > and certainly not make up other questions?   
      
   As you must know, it happens very often on Usenet that someone asks a   
   question which suggests that they have a misunderstanding of the issues   
   involved, or are trying to solve a problem in a suboptimal or impossible   
   way.  In such a case, answering the poster's question directly isn't as   
   helpful to them as trying to get at the root of their problem and   
   addressing other possible solutions.  I thought your question might be   
   of this kind.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)