~kanarip

It was suggested that a quote I used in one of my last blog posts was taken out of context, and inappropriately put on my blog and thus Fedora Planet.

Let me give you a little context here, so that despite the comments cowardly posted anonymously you may make up your own mind.

This thread on the Cobbler mailing list concerns new checks being added to Cobbler for ris-linux and Windows provisioning.

A little background here; In order to notify the user or administrator that implements Cobbler of potential problems with the Cobbler installation, Cobbler has a command: cobbler check. These checks run various little scripts that examine the configuration files, check whether related services are running, and things like that. It also makes a number of suggestions such as additional packages to install for added functionality.

The thread mentioned earlier concerns, amongst other details, whether cobbler check should suggest the installation of the ris-linux package whenever it detects it's not installed, just like it does for cman. I think they ended up deciding that Cobbler should be silent if you configure it to not bother about Windows provisioning but I'm not sure and it's besides the point I'm trying to make here.

When you run cobbler check, you will see that there is several useful messages especially of you have just installed the cobbler package. On your way to resolve and each of every issue raised, you will also find that a number of suggestions are negligible but cannot be disabled or fixed without installing additional packages or manually removing the check from the code. One may not want these extra packages (who needs fencing tools in a bare metal environment, or ris-linux in a homogeneous Linux environment), but nonetheless Cobbler complaints about it. I say "complaints", because that is how the messages are perceived out here in the field.

For diagnostics, this means that a system administrator will run cobbler check to see what problems Cobbler thinks still exist and might cause unsuspected behaviour. Needless to say, you will want the list of messages to be as short as possible.

In my opinion, you then have three options:

1. You add configuration items that tell cobbler whether to even perform the check (i_want_windows: 0), or whether to let the user know about what it finds out (dont_make_suggestions_i_know_what_im_doing: 1)
2. You (silently) check for a package and if it's not installed, you do not check for any configuration items, service or whatever it is you would have checked if the package were installed and you do so silently.
3. You create subpackages of Cobbler that will pull in not only the appropriate checks for said sub-package, but also depends on the package(s) the capability depends upon.

In the first case, I argued that it the list of configuration items to disable just to keep cobbler check as silent as possible might grow and grow over time, so I was ready to suggest some other solutions, thinking I was a good FOSS citizen.

In the second case, Michael DeHaan argued that the check on cman was intended to suggest that the package could be installed to enable the fencing and power management features in Cobbler. That's a good point and there's no arguing about it. Making the checks entirely silent without manual intervention apparently is not a very viable option if you want the output to draw attention to added functionality -so very true.

I on the other hand argued that if Cobbler requires another program to be installed for certain added functionality, it should either (1) require the package though RPM or (2) keep silent and not show messages suggesting you should install it. Of course, not every Cobbler installation wants the cman or ris-linux packages to be installed. If it were installed it might need to be configured as well given the checks Cobbler then runs -although you may never ever need the functionality of said package. Sounds like a waste of time to me if you went this route. Just keeping silent though conflicts with the argument Michael DeHaan made against the second case (the one where he mentions deliberately spitting out suggestions). So, I suggested another solution which brings me to point (3) in the list above.

Now, in the last case where you split up the cobbler package into several sub-packages, one might still want to have yum install cobbler to just install Cobbler with all it's capabilities and checks, and this can be done, and the model for this I suggested in one of my posts is perfectly viable. Many other packages do the exact same thing. But, the suggestion was refused by the technically very sound and apparently well thought through argument that "Now this is just silly. You can live with check output being a few lines longer, I think." -- Michael DeHaan. Let me say that you're right Michael, and that I can live with a few more lines of output, I can. But that coming from someone who says you cannot have distro specific dependencies makes me quote you in blog posts you may wrongfully or rightfully consider be taken out of context and you can live with it, I think.

Coming back to the original point of this blog post -whether I've taken that quote out of context or not. Well, you can make up your own mind. Meanwhile, I've been called a dork, an ass, all by anonymous people of course. Dare you show your face? I know I do!

"No, we can't enforce hard-deps on packages that are not available on all distros."

Isn't that just silly?