Re: New installation program(s)?
On Wed, Jul 14, 1999 at 01:58:22PM -0300, Lalo Martins wrote:
> I realize one of the major discussions here is trashing the
> existing installation system. What follows is an evaluation of
> what would this task mean.
>
> 1: What do we want
> ==================
>
> a: more friendly interface
> a.1: optional GUI install, if X|fb runs on the user's hardware
> a.2: optional no-questions "auto" installation
> a.3: optional networked installation
> where "optional" means "the user may choose it"
>
> b: more detection, less questions
> b.1: use fsresize on the HD if needed
> b.2: use the detection capabilities (ISA, PCI, PnP) of k2.2.x
> b.3: if there is an "install server" running, fetch
> configuration data from there instead of user
>
> c: internationalization - without requiring a recompilation
> c.1: language should be the _VERY FIRST_ question on install
I agree on those goals. We have made a lot of progress on a.1 and c but
other tasks are nearly untouched.
> 2: Steps
> ========
[...]
> DN if install is going to use a "live" fs, where is it;
> first try a CD, then DNI. Should try non-DNI NFS? This
> would require asking the user for address, and, of course,
> doing non-automatic configuration of the network if it
> isn't configured yet)
I've been playing with the idea of using something like HTTPfs.
RW access to the fs is not essential, and an HTTP server is much easier to
setup than an NFS server.
[...]
> NU keyboard layout (is there some way to detect this?)
If you configure the keyboard this late, how may the user answer the
questions before this step?.
[...]
> NDU drivers. Even in "U" mode, we should ask as little as
> possible unless user said something like "expert" somewhere
But the user may need some drivers to partition his SCSI disk or read his
SCSI CD or use his network card. That means he should be able to load
some drivers early in the installation process. Does your proposal
support that?
--
Enrique Zanardi ezanardi@ull.es
Reply to: