On Sat, Mar 25, 2000 at 04:18:22AM +0000, Dale Scheetz wrote: > > It's a judgement call. Therefore I am exercising my judgement. > > > Dependencies have clear specifications. I agree: "The Depends field should be used if the depended-on package is required for the depending package to provide a significant amount of functionality." XF86Setup can provide almost all of its functionality when invoked from within an already running X server. I am willing to entertain the notion of making xf86setup recommend xserver-vga16; "The Recommends field should list packages that would be found together with this one in all but unusual installations." In other words, I accept that many -- perhaps most -- people use XF86Setup primarily to do initial configuration of the X server. The relationship clearly meets the criterion for a Recommends, and clearly does not for a Depends. In so far as there is any gray area here, it *is* within my discretion as package maintainer to decide which way to the split the hair. Since I have to do a 3.3.6-7 anyway, I will make xf86setup Recommend xserver-vga16 specifically. But a dependency is out of the question. > > Also influencing my decision is the fact that, while XF86Setup is a nicer > > looking program, xf86config actually does a better job of generating config > > files. > > Not when it inserts /dev/mouse for my specification of /dev/ttyS0. XF86Setup does worse violence to XKB setups (I gather) for non-US/Candian users. It's far less easy to get XKB set up than the mouse. Therefore, I think xf86config's better chances of getting the Keyboard section right outweigh its perhaps less than stellar handling of the Mouse section. N.B., however that I don't have a bug report or a method of reproducing this xf86config mouse problem. So officially, all I have is your word that you didn't do something stupid. :) > And just what is it? The following has nothing to do with keeping the > VGA16 server off machines that don't need it for a re-configure. If you > include it in common then it gets installed. I already told you I would do this, and I'm furthermore throwing the bone of a Recommends. I'm not inclined to go further that this because I think it is 1) objectively wrong to do so given the definitions in the packaging manual and 2) the level of harassment I'm getting from you on the issue. > > For one thing, XF86Setup does not "depend" on the VGA16 > > server specifically, using the packaging manual's definition of "depend". > > > It refuses to run unless the vga16 server is installed! I'd call that a > strict definition of depends! This is patently false. I just now ran XF86Setup from an xterm. > It refuses to run even when there is another > server installed and that server is set as the default. I can only assume > that it _will_ run if I have a config file. Seems the easier thing to do > is make an initial install into a re-configure by providing a "default" > XF86Config file and "modifying" it to suit the new install. What? There's no way I can ship an XF86Config file that will work for even a majority of situations. If I could, there'd be far less need fot XF86Config file generation tools. Chicken and egg. > You seem focused on the act of using this "setup" program to re-configure > a server, which is very nice, but not the principle use as described by > the package name. This program _is_ expected to work under setup > conditions. And it does. > > Furthermore, users can no longer really use --force-depends with dpkg > > anymore. apt will complain and refuse to function. > > Great! Just what I need, less control over my system ;-( Well, I find that situation less than optimal as well, but you'll have to talk to Jason Gunthorpe about it. -- G. Branden Robinson | <joeyh> oh my, it's a UP P III. Debian GNU/Linux | <doogie> dos it. branden@ecn.purdue.edu | * joeyh runs dselect roger.ecn.purdue.edu/~branden/ | <Overfiend> that ought to be sufficient :)
Attachment:
pgprOCFkniYJW.pgp
Description: PGP signature