[RFC] [PATCH newui] Rework software selection to be driven by comps metadata

Bill Nottingham notting at redhat.com
Wed Jul 25 19:37:01 UTC 2012


Chris Lumens (clumens at redhat.com) said: 
> I know you were posting this hoping to hear about all sorts of problems.
> Sorry, I think it looks pretty good to me.
> 
> My primary concern is a stupid one - do we really want to use
> "installclass"?  That already has a vague enough meaning in anaconda as
> it is.

Yeah, it's just a word that I had around. "Base package set" is the concept,
but that's awfully long to use as the name of it. "Choice"? "Basis"?

> > For any interactive installation, the user will first pick an installclass
> > from the left column. Once they've done that, the right column will populate
> > with the combination of:
> > 1) any groupids from the 'optionlist' for that install class
> > 2) any groups in comps that are marked as 'uservisible'
> > and the user can select any of these to add onto their installation.
> > 
> > The second of these is for some level of backwards compatibility with existing
> > third-party repositories.
> 
> It's always hard to tell certain things from looking at UI patches.
> What does the right side start out as when you first pull up this spoke?

Empty.

> Also, it sounds like someone's going to have to go through comps and
> mark everything as uservisible=False that would appear as an option of
> some class, correct?

Yes, once we get the format nailed down, more drastic comps reorgs can land.

> > This has not been wired into:
> > - pykickstart/kickstart files
> > 
> > Should be doable; need to pick a reasonable character to use as a signifier. ('^'?
> > '$'?)
> 
> I don't have any sense that one character is more meaningful than any
> other.  We can just pick something.  Wiring it into pykickstart should
> be easy, or I've not done my job.

On the yum lists, James suggested just re-using '@' and assuming that there
would never be groups and classes with the same name - that may be
impractical.

Bill


More information about the anaconda-patches mailing list