On Wednesday 14 June 2006 14:33, Wesley Bland wrote:
> Continuing the discussion about Configurator:
> 
> What is your idea about moving the configurator stuff to the ODA?  I'm not 
> opposed to the idea as it would probably make things simpler for me, but the 
> problem that I run into is that I only have 6 1/2 weeks left here and if we 
> decide to go with the database option, I'm afraid that it wouldn't be ready 
> to go in time with all the current effort being poured into 5.0.  The other 
> problem that you run in to when moving stuff into the database is that while 
> it's easy to look up and modify the keys and their values, you will also need 
> some fields to keep track of a few other things related to that key such as a 
> description and possibly a type (if you wanted to use things like radio 
> buttons or check boxes) to know how to group things.

All the applications need for configuration are the variables, maybe the
context. So the minimalistic approach is:

Table  : Configurator
Fields : opkg, name, value, context

"context" is reserved for per-image configuration, for example.

So:
opkg=ganglia
name=gmond_port
value=8649
context="global" (or "" empty string)
would be a valid entry for one variable of the global ganglia configuration.

context="image:myimage"
would specify that the configuration variable is valid only for the image
myimage, but not globally.

This allows us to access the configuration variables easilly, even from the
client nodes or from chroot inside an image. Reconfiguring something will be
easier and backing up the configuration will also be much easier (as all
relevant info lives in the OSCAR database).

The method how we get these variables into the database, how they are
displayed, in which form (html? anything else?) they are collected, should be
unrelated to the way the variables are stored. Actually the variable
collection method should be decoupled from its storage.

So why not describe the collection method in an html or xml file? With the
current configurator.html files this would already work fine (I know this is a
problem for CLI) and be a pretty limited effort. If you can move one step up
and describe the input forms in XML, why not? The XML file can state that a
variable will be collected by clicking a radio-button or selecting something
from a pop-up menu, in the end the variable will have exactly one value and
this is what we'd like to find in ODA.

So I'm an adept of what you call hybrid method. It is because an XML file can
later be easilly integrated into a web gui, but no matter how fancy you make
the GUI, it will not change the way how the variables are stored in the
database.

Best regards,
Erich



_______________________________________________
Oscar-devel mailing list
Oscar-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oscar-devel

Reply via email to