Hi,
Yesterday, I chated with one coder about the database model of GLPI.
Some background : Last year, I discovered IRMA and its database model. I
was disapointed to see that every text input was saved as-is in the
computers/network... tables instead of being saved as references.
I spoke with IRMA's author that replied he knew this was crap, but he
had no time to improve (and as you know, he left this project soon after).
The problem is :
- Either we store every text in the computers table. This is easy, this
is simple, and nothing has to be changed. The problem is that if we
change a single comma in a part table (in the 'os_dropdown' table for
example), the queries we'll do from now won't be reliable (Imagine I
change from 'windowsxp' to 'Windows XP'. The content means the same, but
my queries won't work).
- Either we store *references* in the computers table. The dropdown
tables must have two rows (id and name for example). The queries will
work, we are free to change the labels. This is cool.
This implies a change in the database model. I would be ready to involve
into this sub-project if you wish, but before going any further, I would
be glad to receive some comments.
--
Nicolas Ecarnot
PS : Do we have to speak english on this list ? :o)
