[gentoo-dev] USE database?

[Zach Forrest]

Hi. I've just signed up to the mailing list and wanted to put in my 2^n 
cents. I've just switched over from Debian and am very impressed with 
the Portage system and find administration and configuration to be very 
straight forward and enjoyable. I have learned a great deal about Linux 
in general during the transition (which is almost done -- I think). So, 
without further adieu....

One thing I think needs some refinement is the USE system. I was 
thinking that using a format that is more strict would allow for more 
flexibility and make it easier to automate both the process of 
generating the make.conf file and for tools to administer the USE 
variables. It may also allow for some improvements in the Portage system 
as well.

Using the docs in make.conf as the starting point, I think it would make 
sense to have a simple database format, or, if using python a list of 
dictionaries. For example, one entry might look like this:

use_var={"name":"esd",
          "description":"Enable enlightenment sound daemon support.",
          "priority":"OPTIONAL",
          "depends":"media-sound/esound"}

where:

priority => RECOMMENDED | REQUIRED | OPTIONAL | STANDARD |
             EXPERIMENTAL | DEPRECIATED | DONTUSE | CANNOTLIVEWITHOUT

depends  => packages required to satisfy the USE variable

Note the "depends" entry. This may allow the Portage system to become 
more flexible. Rather than just silently checking for use flags it may 
be nice to let the user know which USE variables are supported and 
possibly give the option to install supported packages, easily located 
through the "depends" field in the "database". This might serve as 
something similar to the Debian "recommended" package option. Maybe 
including a USE or OPTIONAL flag in the ebuild file that lists all of 
the possible USE options could be added.

Also, maybe adding a "--satisfyuse" flag to the ebuild/emerge command.

I know Gentoo is designed for more advanced users and what I have 
described may seem gratuitous, but I think it might make things easier 
in the long run. (Also, even advanced users can do with a _little_ 
convenience.)

It may, then, also make sense to give a little more structure to the 
optimization settings. For example:

optimization_var={
     "host":"i686",
     "chost":"i686-pc-linux-gnu",
     "cflags":"-mcpu=i686 -march=i686 -O3 -pipe",
     "cxxflags":"-mcpu=i686 -march=i686 -O3 -pipe"}

where:

host => i386 | i486 | i586 | i686 | k6 | athalon [ | ppc ]

And, for the sake of completeness, why not the FETCHCOMMAND:

fetch_cmd={
     "name":"Lukemftp",
     "command":"/usr/bin/lukemftp -s -a -o \${DISTDIR}/\${y} \${x}"}

If I recall correctly, my inspiration for all of this was when, during a 
recent update, I noticed that some new USE variables were available. I 
found it annoying to see which ones I had already included (searching 
through my long and possibly overkill USE string) and if there were any 
new ones that I wanted. Then I thought, "Gee, wouldn't a nice ncurses 
interface be great -- a couple of check boxes and then I have more time 
to attend to something only a little more important?"

As this is my first (and hopefully longest) message please give me some 
feedback, both good and bad -- I can take it.

I am thoroughly enjoying Gentoo Linux. Thank you to everyone.

Regards,
Zach Forrest