[Fedora-livecd-list] rfc/suggestion: livecd kickstart format should include url install method

[Douglas McClendon]

Jeremy Katz wrote:
> On Mon, 2008-01-14 at 13:42 -0600, Douglas McClendon wrote:
>> Jeremy Katz wrote:
>>> On Sun, 2008-01-13 at 22:32 -0600, Douglas McClendon wrote:
>>>> The main problem, is that no install method is specified.  For now, I'm 
>>>> just going to do a little parsing, taking the first repo line, and 
>>>> converting it into
>>>>
>>>> "
>>>> install
>>>> url --url=<bla>
>>>> "
>>> While not required, this syntax should work today.
>> Then my suggestion is mainly about what the example kickstarts look like...
>>
>>    The url syntax
>>> doesn't support as many things like mirrorlists, though, and thus isn't
>>> as preferred.  See the methodToRepo bits
>> I noticed this as well.  The intuitive answer would seem to be that that 
>> the url syntax should support mirrorlists.  Is this planned for F9?
> 
> That means the loader would need to have support for mirrorlists.  And
> proxies.  Etc, etc.  The better answer is going to be the work which is
> ongoing for Fedora 9 so that we can define repos at stage2 time... at
> which point, we could just use a config with repos for install-time
> also.

Thats cool- it was the functionality and compatability I wanted, and it 
sounds like that is the plan.

> 
>>>> Other than that, the only other thing I think I noticed was that it 
>>>> failed to go completely automated due to lack of specified root password 
>>>> entry.  Again, I'd suggest explicitly requiring that entry.
>>> An empty password is going to cause problems, though, and that's what we
>>> end up wanting for general live image use.  And if you want to have a
>>> root password, you can specify it and then things will work in either
>>> case
>> I'd suggest in this case supplying the rootpw entry, and then using 
>> %post to remove the root password.  Again, the idea being that the 
>> example kickstarts supplied should work well as non-livecd kickstarts.
> 
> This just makes things convoluted, though and harder to understand, not
> easier.  Because now to figure out why there's not a password, I have to
> go and parse through the scripts.  Which sucks IMHO

IMHO it sucks to have a kickstart which wouldn't run in an automated 
fashion in anaconda, and I don't see how the process of figuring out 
there is no password by the non-existence of any configuration text is 
any better.  But not a big, or even a small deal.

-dmc