Subject: Re: implementing a name service switch...
To: Luke Mewburn <lm@melb.cpr.itg.telecom.com.au>
From: Chris G Demetriou <Chris_G_Demetriou@LAGAVULIN.PDL.CS.CMU.EDU>
List: tech-net
Date: 12/05/1994 20:57:07
[ why is this on tech-net?  it doesn't belong here.  however, now that
  it is... ]

> Basically, a config file such as /etc/nsswitch.conf (for name
> compatibility with Solaris, HPUX, and probably every other SVR4) is
> required. Then, for each database such as passwd, group, netgroup,
> hosts, services, etc, there is an entry describing the lookup
> procedure from various sources such as local (files), bind, yp,
> or any other transport mechanism we provide.

My personal cut on this is that it's not worth the extra complexity it
adds.  Add that to the fact that i've yet to have a _positive_
experience with such configuration files.  I don't really think it's a
good idea.  (in particular, most implementations of this switch that
i've seen fall over badly, if the file isn't present...)

I also don't think it's worthwhile putting HESIOD code in the source
tree.  I actually am marginally against YP, but a large number of
people actually use it.  I see adding HESIOD as adding an awful lot of
code-maintenance overhead, for a very, very small gain.



You'll note that some of the functions that YP support has been added
for have required significant hacking to get 'right' in the presence
of YP.  adding HESIOD adds another variable that needs to be dealt
with.  adding complete configurability adds another one (At least!).



chris