On 07/29/2015 05:13 PM, Martin Babinsky wrote:
On 07/29/2015 01:25 PM, Jan Cholasta wrote:
Dne 29.7.2015 v 12:20 Martin Babinsky napsal(a):
Initial attempt to implement

Some points to discuss:

1.) name of the config entries: currently the option names are derived
from CLI options but have underscores in them instead of dashes. Maybe
keeping the CLI option names also for config entries will make it easier
for the user to transfer their CLI options from scripts to config files.

NACK. There is no point in generating config names from CLI names, which
are generated from knob names - use knob names directly.

The problem is that in some cases the  cli_name does not map directly to
knob name, leading in different naming of CLI options and config
entries, confusion and mayhem.

These are some offenders from `ipaserver/install/server.py`:

On the other hand, this can be an incentive to finally put an end to
inconsistent option/knob naming across server/replica/etc. installers.

If the names are different than cli names, then they should be made discoverable somehow or be documented.

2.) Config sections: there is currently only one valid section named
'[global]' in accordance with the format of 'default.conf'. Should we
have separate sections equivalent to option groups in CLI (e.g. [basic],
[certificate system], [dns])?

No, because they would have to be maintained forever. For example, some
options are in wrong sections and we wouldn't be able to move them.

I'm also more inclined to a single section, at least for now since we
are pressed for time with this RFE.

That's not to say that we should ditch Alexander's idea about separate
sections with overrides for different hosts. We should consider it as a
future enhancement to this feature once the basic plumbing is in place.

3.) Handling of unattended mode when specifying a config file:
Currently there is no connection between --config-file and unattended
mode. So when you run ipa-server-install using config file, you still
get asked for missing stuff. Should '--config-file' automatically imply

The behavior should be the same as if you specified the options on the
command line. So no, --config-file should not imply --unattended.

That sound reasonable. the code behaves this way already so no changes

There are probably other issues to discuss. Feel free to write
email/ping me on IRC.

(I haven't looked at the patch yet.)

Please take a look at it ASAP. I am on PTO tomorrow and on Friday, but I
will find time to work at it in the evening if you send me you comments.

Petr Vobornik

Manage your subscription for the Freeipa-devel mailing list:
Contribute to FreeIPA: http://www.freeipa.org/page/Contribute/Code

Reply via email to