[Freeipa-devel] kra an ca instance installation

[Jan Cholasta]

On 25.8.2015 15:37, Simo Sorce wrote:
> On Tue, 2015-08-25 at 11:34 +0200, Jan Cholasta wrote:
>> On 25.8.2015 11:25, Jan Cholasta wrote:
>>> On 24.8.2015 19:51, Simo Sorce wrote:
>>>> Why do we have cainstance.py and ca.py and krainstance.py and kra.py in
>>>> ipaserver/install when you always need both files to do anything around
>>>> installation of the ca ?
>>>>
>>>> Is there a motivation ?
>>>> Or can I simply provide a patch to remove the ca.py and kra.py files an
>>>> unify all code in the proper *instance.py file ?
>>>
>>> ca.py and kra.py are the proper files ready to be migrated to the new
>>> install framework, cainstance.py and krainstance.py will be removed.
>>
>> ... once the migration is done. (Hit send button too fast.)
>>
>> The motivation is that *instance.py do not provide a uniform interface,
>> have a lot of redundant and duplicate stuff and are generally unfit for
>> any further extension.
>
> I have been changing only the instance files, so we are going in
> different directions.

I don't see how, {ca,kra}instance.py code is called from {ca,kra}.py. 
{ca,kra}.py contains all the code that is actually needed to install 
CA/KRA that is not in {ca,kra}instance.py and was previously scattered 
around ipa-{server,replica,ca,kra}-install.

>
> I do not really care what file we are going into, but there is a lot of
> code in the installer now that does not tell the user a step is being
> done, while instances do that through the step interface.

This code was always there, we only moved it to a single location. See 
git history.

>
> The step interface is also a very good way to let someone that read the
> code see what is going on and follow each step.
>
> Are you proposing to stop going through the instances steps ? I found
> the current way kra and ca installation is setup basically a regression,
> it took me a *lot* longer that it should be needed to follow through all
> the steps that are really taken.

Again, we only moved the code around, and I don't think we created any 
regressions.

>
> Simo.
>


-- 
Jan Cholasta