On Wed, Dec 07 2016, J. Bruce Fields wrote:
> On Fri, Dec 02, 2016 at 02:58:27PM +1100, NeilBrown wrote:
>> This is an RFC series. A little voice at the back of my head keeps
>> telling me that I'm over-engineering, but there isn't really that much
>> new code, and I think the result has a lot to recommend it.
>>
>> But please tell me if I'm wrong.
>>
>> - Various daemons (not all) are enhance to accept configuration
>> information from /etc/nfs.conf
>> - the conffile reader is enhanced to support include files, and
>> particularly to be able to include /etc/sysconf/X or /etc/defaults/X
>> files usefully
>
> Currently those files are actually sourced by a shell, right? So in
> theory people could be doing tricky things in there that would no longer
> be supported. Probably unlikely, though, OK....
They are sourced by a shell, but they are read and written by tools.
SUSE has YAST, Debian has debconf. I assume redhat has something
similar.
Those tools might let some shell syntax through, but people who try
games like that are already taking a risk I think.
>
>> - nfs-config.service is removed, because it isn't really needed with
>> the above.
>> - documentation for all the above is provided, including a new
>> nfs.systemd man page which gives the bigger picture.
>
> Still looks pretty good to me.
>
> I'm a little worried about user interface churn. We're not done yet
> explaining that people have to run nfs-config.service after changing
> things, soon we'll start telling them oh, never mind about that and oh,
> by the way, you may want to start migrating your configuration to
> /etc/nfs.conf....
Did we tell people to run nfs-config.server? The intention was that it
wouldn't be needed. Any systemd transaction that needed nfs
configuration should run the command once.
Commit: c4940fad2a73 ("systemd: ensure nfs-config service is re-run as needed.")
should have made that happen.
Yes, some migration is needed. I see distros as the primary target for
that, and the 'include' and '$name' functionality is supposed to allow
that to be transparent. i.e. the configuration stays as it is, it just
gets into nfs daemons by a different path.
If people want to edit /etc/nfs.conf directly, then they are on their
own, just like people who edited their systemd unit files directly,
or created replacements in /etc.
Thanks,
NeilBrown
Attachment:
signature.asc
Description: PGP signature
