Re: [gentoo-user] chicken <--> egg (NFS & tty video)

[Alan McKinnon]

Apparently, though unproven, at 18:01 on Saturday 14 May 2011, Indi did opine 
thusly:

> On Sat, May 14, 2011 at 05:53:56PM +0200, Alan McKinnon wrote:
> > Apparently, though unproven, at 16:37 on Saturday 14 May 2011, Indi did
> > opine
> >
> > thusly:
> > > True, just be aware that if you enable gtk *globally* you will end up
> > > building the gtk interface for absolutely everything which has that
> > > option.
> > > Far better (IMO, YMMV) is to use /etc/portage/package.use specify such
> > > things per package. Unless, of course, you like having a gtk GUI for
> > > everything.
> > >
> > > :)
> >
> > No, it is much better to enable such a flag globally and *disable* it
> > using package.use where you do *not* want it.
> >
> > Personally, I have better things to do than examine every new or changed
> > package that shows up after avuND world and edit package.us for every
> > single flag in that huge list.
>
> Sounds like the old "6 of one, a half-dozen of the other" to me...
> What makes the subtractive method better?

It's not subtractive as disabling a flag globally and enabling it when needed 
is the same thing negated.

I'm pointing out that by their nature, most global USE flags are exactly that 
- intended to be global, especially those in use.desc. For the most part the 
user will want the support they provide to be global. When that is not the 
case (the lesser case), an option exists to override the global setting in 
package.use

What you proposed is that one never use global flags and always enable/disable 
them package by package. That gets really tedious with flags used in many 
ebuilds, such as USE=gtk.

Abstaction is good, leverage it to gain the benefits when it works in your 
favour.


-- 
alan dot mckinnon at gmail dot com