1.618: How can I hack libc.a to alter how hostnames are resolved?

[ 
  Editors note:  The implications of corrupting libc.a should not be
  understated.  While recovering from a corrupt libc.a should be a
  matter of booting from floppy I wouldn't wish that on anyone with a 
  supervisor or users to support :)

  Before you attempt this,  you might want to read 2.07 first for
  advice on recovering from a destroyed or corrupted libc.a.
  Remember:  backups are a VERY GOOD IDEA.

  enough of the weak kneed quivering...
]

Nate Itkin (a colleague) is the individual who did the initial work.
I suspect he would rather not be troubled with any questions you may
have in this area and I can probably answer them just as easily.  The
motivation for this was to use DNS for MX records (we are using IDA
sendmail) but NIS for most hostname lookups.

Secondly, warnings:  IBM has not given any sort of stamp of approval on what
we have done (though they appear to have integrated resolv+ into AIX 4 based
on some release notes I saw).  While we have been using it successfully for
about one year and have seen no problems, we can't and won't promise the same
for you.  This procedure plays around with libc.a which, as you probably know,
is an XCOFF shared library under AIX and hence a critical part of the running
system.  You assume all the risk if you try to install this.  All the usual
disclaimers about liability, etc. apply - there is no warranty associated with
any of this.

Furthermore, know that any PTF you subsequently install that updates/replaces
libc.a implies that you have to execute most of these procedures again.

The details on how to use resolv+ under AIX 3 (it's only tested under
3.2.5 but should work for all point releases), have been moved to
section 8.09.