Dave Miner wrote:
> Kyle McDonald wrote:
>> Which suggestion?
>>
>> Did the old installer create a JumpStart profile and then call 
>> pfinstall?
>>
>
> Yes, it did.
>
Oh wow. I didn't know that. Why didn't it leave the profile around where 
it could be reused, or if it did why wasn't that documented?
>> While a hared library could eliminate most of the duplication between 
>> an interactive installer, and an Automated one.  It seemed like 
>> (given that one requirement of the interactive one was to ask all 
>> questions upfront befroe starting the install) having the interactive 
>> GUI create an AI profile, and call AI, would also eliminate 
>> duplication (and therefore QA and bug-fixing costs) Not to mention it 
>> would actually 'test' the profile generated by using it.
>>
>
> A simple installer such as the GUI isn't going to test anything 
> interesting; we'll write real test cases for testing.  In general, 
> yes, any install path will be generating an internal representation 
> equivalent to an automated installation profile and using the 
> appropriate interfaces from there; they'll mostly be common.
I agree that the simple installer isn't going to create complicated 
profiles. What I meant by 'testing', is that the user would know without 
a doubt that the profile left behind by the installer would definitely 
recreate the install, because it *had actually been used* to do the 
original install.

There's a difference between saying:

 "Here's what I used to do the install, use it if you want to repeat it."

 and

"I used something thing else, but this should recreate it if you need to."

  -Kyle


Reply via email to