Greg A. Woods wrote:

> [ On Saturday, December 4, 2004 at 15:11:47 (-0800), Gerald Richter wrote: ]
> 
>>Subject: Re: sysinst problems
>>
>>just sometimes I 
>>need an expert knob so I can help the OS around some problem bit of 
>>hardware that works most the time, except when it does something the 
>>installer doesn't expect.
> 
> 
> Well, I do agree in principle, even though I may not have made that fact
> as clear as I should have.

If you agree in principal, why are you arguing against your principals? O_o

> I think the point though is that "sysinst" is just a tool-using tool
> itself and _nothing_ it does can ever prevent anyone, expert or
> otherwise, from using the underlying tools in some different way for
> whatever reason.  Indeed "sysinst" even has a feature to allow it to
> write out in shell-script form the very commands that can be used to
> duplicate its every action based on the current session's user input.

That's all and good and mostly useless.

>>sysinst doesn't believe in having an expert knob.
> 
> 
> Nor should it, ever.

Every installer should always. Just to put it in equally rabid zealot 
terms. Hehehe. Actually, not all installers do enough to _need_ an 
expert knob, most don't.

> Everything it does can be done by hand with every variation imaginable,
> even if it does take a little bit of imagination or deeper understanding
> of what's really going on under the hood.

Who says I wanna do the whole thing by hand? I just wanna twist enough 
of the knob so that it can get around a problem and then let it do it's 
thing. Why should I manually do everything when there is this elegant 
program that will do it for me, except for while it's barfing on this 
problem that it won't let me help it with because it doesn't have an 
expert knob??

[more rantage removed]

> The effort to which someone has to go to "break the rules" should depend
> on the importance those rules have on the correct, safe, and secure
> operation of the system.  Disklabel information is incredibly important
> to the operation of the system and since it's only read once and then
> trusted for almost everything else the system does with the storage
> device at the lowest levels it's wise for the "system" to do as much to
> validate and verify that information as is possible before it is
> trusted.

I would beg to differ on that. Yes, its important, but so are a lot of 
other things. What is ultimately harmed if we have an expert knob that 
allows us to ignore the fact that the sysadmin who obviously should have 
as much rights to the system as the kernel (remember, the sysadmin is 
trusting the kernel to handle the second to second tasks in his 
computer) feels like setting up something that _looks_ bizarre, but 
_might_ work given circumstances outside our programs range of 
comprehension? There's no saying the knob has to be easy to find 
(documented in any way) or activate (make 'em spell out WTFYOUIDIOT in 
binary on the command line to activate it), but it _SHOULD_ be there.

> At the moment the kernel trusts the label almost implicitly (so long as
> its basic values don't violate each other drastically) since presumably
> only someone in a position of trust can write that information to the
> disk.  Therefore the sanity checking needs to be implemented in the
> tools that are used to create and write the label.  At the moment I see
> the different levels of sanity checking done by sysinst and then by
> disklabel to be quite appropriate to their intended usage.  That's why I
> think this cry about sysinst being to strict is _way_ out of line.

Note: Person installing is almost guaranteed to be a person in position 
of trust. The matter of if they should have the trust is far too 
debatable for this thread. And if they screw it up it's their fault, not 
that of sysinst... Particularly if they twisted the impossible to find 
and activate expert knob.

> The problem with the hand-holding done by some other commercial systems
> is that they don't often let the expert work under the hood.  This is
> where the software tools-based philosophy of unix-like systems is a big
> feature amongst experts.  I.e. it's not even remotely fair to compare
> their rigour with that of "sysinst" since we already have a plethora of
> "expert friendly" alternatives to "sysinst" and always will have.

Unfortunately you are arguing for _EXACTLY_ that level of hand holding 
in sysinst. That's a problem. I know what sysinst does, I could do it on 
my  own, but why should I have to??? It's a perfectly good tool at doing 
what it does, save for it does occasionally need an expert knob to help 
it around a problem, or something bizarre that I happen to be doing but 
would like it's assistance with so I can go to lunch while it does things.

Anywho, enough of my blathering. I'm finding your arguing very amusing, 
because it goes so totally against the grain of what I've come to expect 
from persons connected to open source projects.

Have a good one all,
--Gerald