* Dave Miner (Dave.Miner at Sun.COM) wrote:
> Glenn Lagasse wrote:
>> Hey Clay,
>>
>> Thanks for the feedback, my thoughts inline.
>>
>> * Clay Baenziger (clayb at sun.com) wrote:
>>> Hi Glenn,
>>>     I'm all for the simple approach. I think the power of the DC is for 
>>> a user to twiddle the bits on and off and extend only if necessary 
>>> and  skilled enough.
>>>     If you have a finalizer which does the checksums then someone 
>>> wanting to extend this functionality can always easily slurp-up and 
>>> modify that finalizer script. Easier is certainly better!
>>
>> And why can't they also slurp up and modify the existing finalizer
>> scripts that contain the new checksum functionality?
>>
>>>     Though I agree with the simplicity of Karen's desire to have the   
>>> create_iso and create_usb steps include the hash generation, having 
>>> it all in one place (a single finalizer script) makes a modification 
>>> by skilled users easier without really any added complexity for a 
>>> novice.
>>
>> I don't know that I see your argument here.  Having a distinct finalizer
>> script that just computes hashes vs computing hashes inside the
>> finalizer scripts that actually create the images you want hashes for
>> doesn't seem to be any different to me in terms of ease of modification.
>> What exactly is easier about modifying a finalizer script that only
>> computes hashes?
>>
>> By including the checksum generation in the existing image type
>> finalizer scripts (by image type I'm referring to the create_iso and
>> create_usb scripts which are the only two types of images DC can
>> construct at the moment that you'd want checksums for) I'm linking image
>> hash generation with the actual image construction which just seems
>> natural and logical to me.  Moving the hash generation out into a
>> finalizer script of it's own buys us the ability to pause and resume
>> around that script but I don't find that to be of any significant value
>> considering how simple of an operation generating the image hashes are.
>> It isn't a long-lived complex process so I don't know that there's any
>> 'visibility' gains achieved by being able to stop construction around
>> generating the hashes.
>>
>
> If the hashing were configurable, the advantage would be to be able to  
> re-run with different hashing without regenerating the images.  However,  
> since you don't seem to be implementing configurability of the hashes,  
> then I guess that's moot.

That gives me something to think about.  I think I had this scenario in
the back of my head, but wasn't mentalizing it when I was coming up with
this.  I think I can see how and why having the separate finalizer
script (in conjunction with configurable hash types) would be desirable
more clearly.

Let me go ponder a bit.

Thanks Dave.

-- 
Glenn

Reply via email to