Re: [GSOC 2008] update: install-tool

["Zach Wegner" <zwegner%gmail.com@localhost>]

On Tue, Jul 15, 2008 at 12:06 PM, David Laight <david%l8s.co.uk@localhost> 
wrote:
> Indeed, but it is rather important to be able to query the user to fix
> the input and then retry set installation.
>
>
>        David
>
> --
> David Laight: david%l8s.co.uk@localhost
>

I would agree. But it's not such an easy problem to tackle:
1. Calling the back end from the front end to check would be cleaner
IMO in design, but it's not perfect. That is, you can test something
as much as you want in the front end, but it could still fail in the
back end.
2. Calling the front end from the back end is a little uglier IMO.
Mostly just aesthetic reasons though, but it also doesn't allow for
testing when you do mass installs.
3. I think the best solution would be to have both, in a way. You
allow the checking feature in the front end, and then sysinst becomes
a loop that repeatedly calls the front end and then the back end until
installation is successful or the user gives up.
4. Regardless of what is chosen from the above, it doesn't solve the
problem of unattended installs. Ideally it should be possible to
install without a monitor or a keyboard if everything is set up right.
But if an FTP site is down, it should have some sort of error
correction. My idea is to have backup media, so that for when a set
fails on CD, we can try NFS, and if that fails, FTP, etc. The
implementation shouldn't be too hard, as each type of media specifies
what sets it's installing, and it will just try all of them and take
out any duplicates. You can also specify multiple instances of each
media type, so you can try different FTP mirrors, etc.

How does that sound, everyone?