On 17 Jul 00:57, Dominique Chabord wrote:
> Le 16/07/2014 15:29, Cédric Krier a écrit :
> > On 16 Jul 14:43, Dominique Chabord wrote:
> >> Hello,
> >> I think this option is required and should be defined per database and
> >> not in configuration file.
> > 
> > Soon it will be the same as one database per process.
> 
> This is not a valid argument. If Tryton runs a single database at a
> time, it can run several databases one after the other.

Don't understand.

> Nevertheless
> will Tryton be single company too ? because this parameter is company
> dependant.

Tryton is already mainly single company, it is a fact see previous
discussion about multi-company.
I don't see any advantage to have multi-company if it is to have
everything different.
Here, I just apply the KISS principle.

> > Also I don't want to have it in the database for many reasons:
> > 
> >     - can not really be changed
> 
> There are many things in the database which cannot be changed and if it
> cannot be changed, it enforces the fact that it has nothing to do in the
> start configuration of the process. What about a module ?

So you want to create a module for just setup a flag?

> >     - will cost too much to read each time
> 
> Then replace the database by a big config file, and you will be fast ;-)
> Out of curiosity, why do you need to read each time from the database
> and one time from the config file ?

Because configuration is by definition static. Data in the database are
not.

> And why is it different for other
> company-related data ?

What have you in mind?

> > There will be more and more things going in the configuration file like
> > the all ldap_connection module, the price digits etc.
> 
> price digits and sum ordering are given accounting parameters of a
> company. Two companies need different values.

If two companies need different values, they should be managed by
different instance.
Once again, mutli-company is a dream. Nobody ever achieve to make it
works out side of simple case (like subsidiary). If companies don't
share the same rules, they have nothing to do in the same database.

> >> Btw, if implemented in conf file, how would the history behave when
> >> conf is changed ? or when database is restored ?
> > 
> > People should be careful to not shoot their own foot.
> 
> tryton must be manageable too.

Exactly, that's why all those stuff should be in a configuration file.

> For restoring any archive of a database
> you will need to know about the configuration file of the original
> server, adapt some parameters on your local server and start it.

Yes and you will also need to know the version of the source code
executed.

> The configuration file should define parameters related to the execution
> context and should not depend on the database you restore.

That's wrong. You can not manage data independently of the code, they
are the same.

-- 
Cédric Krier - B2CK SPRL
Email/Jabber: [email protected]
Tel: +32 472 54 46 59
Website: http://www.b2ck.com/

Attachment: pgpK23PVTJ2pe.pgp
Description: PGP signature

Reply via email to