[Fusioninventory-devel] new generic injection feature

[Guillaume Rousse]

OK, I think I found the user of this hack :)

Le 06/06/2011 14:33, Sébastien Dagnicourt a écrit :
> Hello,
> Of course I disagree a bit with that ... ;)
> First, the ByHand "workaround" is documented. A little only, but it is.
Do you really consider comments buried in the perl module a
user-targeted documentation ? I don't.

> Second, the goal was to provide an easy way for users to add software
> scripts. I mean, as an ocs/glpi administrator, I don't have ( and I
> don't want) to check or modify the scripts that users deliver to me.  I
> just have to copy them at the right place (or better : users just put
> them at the right place).
> Even if the XML way permits more things than software discover, if
> ByHand is removed,  I will have to:
>  1) Inform users where they have to write the output of their scripts
Anywhere, that's the point of using a command-line switch instead of an
hard-coded location.

>  2) Check that the output is correct
It's far more easier to check a file content, wether generated by a
script, or manually produced, than a script supposed to produce this
content. Especially when this script has been written by someone else
without any coding style reference, and you don't have access to the
host where this script is supposed to run...

>  3) Change all scripts already in place.
> 
> And I don't to have this work :)
Why upgrade then ? If you're happy with 2.1.x branch, you don't need to
switch to 2.2.x one.

Of course, I'm interested about any alternative proposal, but:
- it should not be software (or any other category) specific
- it should not run untrusted external code, especially as root user
- it should be documented at user-level
- it should be tested (by some kind of automated regression test, not
just user-approved).

-- 
BOFH excuse #300:

Digital Manipulator exceeding velocity parameters