This post is about some issues with the various inetd packages in etch
(and unstable). This is a case where I think some coordination
between all the packages or some inetd package policy would make them
all generally more usable.
The currently available inetd packages, and a summary of their state:
----------------+---------------------+-------------------------------------
Package | IPv6 support | Source quality / comments
----------------+---------------------+-------------------------------------
inetutils-inetd | tcp = tcp4 and tcp6 | OK, but upstream quiet
| |
micro-inetd | Partial | Not a proper inetd replacement
| |
netkit-inetd | No, and it will be | Terrible. It won't even build
| difficult to add | from the .orig.tar.gz, and the
| | tarball is a mess. IMO, should be
| | removed from Debian.
| | 102 outstanding bugs.
| |
openbsd-inetd | tcp=TCP4, tcp4, | OK. Bug in reload (#382404).
| tcp6, tcp46 | Bug in restart (#376716).
| |
| |
rlinetd | Yes | Not a drop-in replacement.
| |
superd | No | Unmaintained. Candidate for
| | removal?
| |
xinetd | tcp4, tcp6 | Good, but not currently a drop-in
| | inetd replacement (but could be
| | configured to do so).
----------------+---------------------+-------------------------------------
Outstanding issues
------------------
* There is no inetd virtual package, so multiple daemons may be
installed, all using the same configuration file. Is this a use
case we really want to support? Are there really setups running
multiple inetds for a good reason? Having a virtual
"internet-super-server" package or similar with appropriate
dependencies would make them rather more interchangeable, as for
e.g. mail-transport-agent.
* There is no common init script name. Same problems as above.
* netbase only depends on two inetd packages (openbsd-inetd and
netkit-inetd; a virtual depends plus a default would be nicer.
* netkit-inetd
- No upstream.
- Last maintainer upload was 22 months ago. The last three uploads
were NMUs.
- It doesn't build from the original source.
- The original source is a horrible mess, with code duplication (the
source tree has a duplicate copy embedded within itself), and i386
ELF object code and binaries code in the tree.
- The C source itself is not very nice.
- Is this really fit to keep in Debian? It might be better to
remove it entirely given its terrible state.
* openbsd-inetd is the only drop-in replacement at this time
- The other packages have different init script names or need some
work on the package dependencies (e.g. inetutils-inetd). xinetd
is in the same situation, but also needs some work on update-inetd
before it will be suitable as a replacement.
* IPv6 transition
- Should individual packages be made to listen on both tcp4 and tcp6
sockets, or should this be done by the inetd itself, or even
update-inetd?
- Some inetds automatically listen on v6, whereas others need it
explicitly enabling. What should "tcp" vs "tcp4" vs "tcp6" (and
the same for udp) imply?
- Some general policy would be useful here to make the behaviour
consistent and to make IPv6 support as painless as possible for
users.
* Upgrade from sarge and earlier
The inetd daemon installed by default:
etch: openbsd-inetd | netkit-inetd
sarge: netkit-inetd
woody: netkit-inetd (netkit-base, split from netbase)
potato: (in netbase)
slink: (in netbase)
Users upgrading from woody or sarge to etch will not be switched to
openbsd-inetd, whereas new installs will use it by default.
Removing netkit-inetd from the netbase depends should permit this,
but a complete removal would perhaps be the best option.
While it's probably too late to fix up update-inetd and all the inetd
integration issues for etch, fixing the netbase dependency and
eliminating netkit-inetd is doable.
Any thoughts or comments?
Regards,
Roger
--
.''`. Roger Leigh
: :' : Debian GNU/Linux http://people.debian.org/~rleigh/
`. `' Printing on GNU/Linux? http://gutenprint.sourceforge.net/
`- GPG Public Key: 0x25BFB848 Please GPG sign your mail.
Attachment:
pgpiSnlQ7P9Rj.pgp
Description: PGP signature