-----BEGIN PGP SIGNED MESSAGE-----


>>>>> "Chris" == Chris G Demetriou <Chris_G_Demetriou@ux2.sp.cs.cmu.edu> writes:

  [arg. Who said this? Gordon Ross?]
    >> Actually, that is exactly the reason I have been lobying for an
    >> implementation of the nsswitch code that removes all the direct
    >> references to the YP stuff, etc. from libc (and using rtld to
    >> load those when needed).

    Chris> Unfortunately, any way you look at it, that's quite likely
    Chris> a step backward on systems which don't support shared
    Chris> libraries, because if they don't support shared libraries
    Chris> they're not very likely to support rtld and dynamic loading
    Chris> of those modules.

  As someone who went through the horror of trying to
de-/etc/netconfig the -lsocket,-lnls stuff in SVR4 while porting a
firewall to UnixWare 2.01 (ick. even getcwd() had an obvious bug in
it), I find I must agree with Chris. Yes, I did have system source.
    
  /etc/nsswitch.conf is okay. Not wonderful, but I can mostly deal
with it. It doesn't list .o files to link in.

  [warning: slipperly slope argument comming up. I think NetBSD has
more self-control than that, btw]
  It may lead to other things, e.g. /etc/netconfig. Rather than worry
about who has DNS and who doesn't have DNS, etc... why not just make
DNS easier to configure, write "basic.named" to read /etc/hosts and do
away with the rest? 
  I think Sun has realized this, and wishes to make their "doors" the
way to do most of this. I'd say "doors" are to /etc/nsswitch.conf what
sockets are to TLI. >Best Current Practice<

  I am one who rather thought the the method of getting DNS on SunOS
(rebuild the shared library) was the *MOST* elegant solution. I don't
run NIS, no NIS. I do run DNS. Add DNS. It isn't like I need to do
this at runtime. This is where /etc/nsswitch.conf and /etc/netconfig
and even lkm fail: I DO NOT NEED TO DO THIS ON A DAILY BASIS.
  SysV with its /etc/conf and idbuild system was very nice. The fact
that it was required and one could not build a "GENERIC" kernel was
bad (idbuild in a firewall? Yuck), but not unreasonable. It's bigged
fault was that you needed a compiler.

   :!mcr!:            |  Network security consulting and 
   Michael Richardson |      contract programming
 WWW: mcr@sandelman.ottawa.on.ca. PGP key available.
  
  
  

-----BEGIN PGP SIGNATURE-----
Version: 2.6.3ia
Charset: latin1
Comment: Processed by Mailcrypt 3.4, an Emacs/PGP interface

iQBVAwUBMq5OLNTTll4efmtZAQEZLAIAoNYY3LR+NHcSNQ2Sae7iRhS8VsP26R0z
zfdVIY6Zfg2er9v9iGpuQhxI+dHCp531zpiTSU9XlmBl18YfxoX31Q==
=iUDe
-----END PGP SIGNATURE-----