Re: inetd Enhancements

[Mouse <mouse%Rodents-Montreal.ORG@localhost>]

> I run servers on non-standard ports (doesn't everyone?)

No.  My inetd.conf has no standard servers on non-standard ports - I
just checked my main desktop machine and my main house router.  (They
have non-standard servers on non-standard ports, but that's the use
case I mentioned of wanting to run a non-standard service in the first
place.)

I'd hazard a guess that fewer than half the people in the discussion,
even, would want to fiddle their service definitions under dholland's
scheme, and we're probably extreme outliers in the user population.

> One could certainly create a whole new syntax for inetd.conf (or some
> replacement in any format) but it needs to be able to specify almost
> all the data that is currently there, one way or another (whether a
> dgram server is wait or nowait is probably the one thing that cannot
> sanely be altered, and depends upon the implementation of the server,
> and its protocol).

And whether a server is dgram/udp or stream/tcp.  For most daemons, for
most sites, fiddling the run-as user or the wait/nowait setting is also
ill-advised at best.  And that's four of the seven fields (counting the
whole command as a single field, and the wait/nowait and rate-limit
field as two fields, because that's really two settings smushed into a
single field).  (The other field is the port number.)

No, I think dholland is basically correct.  Like dholland, me, and
most of the NetBSD developer population, you are an outlier among
outliers in the extent to which you (we) want to do unusual things with
your (our) systems.  (We are also outliers in our competence to do
unusual things and make them work right.)

If done right, the syntax could even allow overriding fields in the
inetd.conf.  Instead of just

ntalk on flags -l
ntalk off

you could have something like (strawman alert)

ntalk on port=12345 flags -l -q magic

to use the ntalk definition but with the port setting overridden.

With a little more care, you could even have it recognize the existing
syntax and interpret it compatibly, because in the existing syntax
every line either (1) is empty, (2) is a comment, or (3) begins with a
port name or number.  This offers lots of room for compatability and
extensibility.  Recognizing that the second field in existing
non-comment lines is always either stream or dgram offers even more
flexibility, as long as you ensure that the second field in a new-style
definition is never either of those.

/~\ The ASCII				  Mouse
\ / Ribbon Campaign
 X  Against HTML		mouse%rodents-montreal.org@localhost
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B