Am 01.11.2011 15:01, schrieb HAT:
> Hi,
> 
>> ExecStartPre=/bin/systemctl start netatalk-cnid.service
>> ExecStopPost=/bin/systemctl stop netatalk-cnid.service
> 
> This is very convenient because cnid_metad is started/stopped
> simultaneously with afpd.
> Almost all users use "cnidscheme:dbd".

well so use a shellscript here which reads the configuration
and starts/stops cnid depending on this

> However, this is not strict.
> If cnidscheme is "tdb" only, it's not necessary to run cnid_metad.
> strict setup is:
> After=netatalk-cnid.service
> 
> Which should we choose between convenience and stringency?

daemons like netatalk are not easy to handle here but the
current syystemd-unit from the sources is a bad joke because
it is useless starting only a shellscript which fires
up some processes or not since systemd does not know
anything about the real processes

there is no benefit compared with the old sysv-script

> In the future, the situation will become more complicated.
> A new daemon "netalockd" is added to netatalk 3.0dev

currently i see a big design mistake fire up different
daemons which may or may not depend with shell-magic

take a look at postfix with his master.cf and master process
to see a perfect multi-process daemon starting only configured
processes and most of them only on-demand if they are needed

however, my needs are fullfilled with afpd/cnid and i know how
to overwrite the generated sytemd-unit with my own in the
rpm-spec-file and my primary intention was to point out
that the current systemd-unit is horrible broken by design




Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
systemd-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/systemd-devel

Reply via email to