On 01/14/10 10:17, Eric J. Ray wrote:
> On Jan 14, 2010, at 10:30 AM, Sarah Jelinek wrote:
> 
>>
>> jeanm wrote:
>>> From reading both Dave and your response I think the confusion is due to 
>>> what we consider a checkpoint. I consider a
>>> checkpoint a single place in time, that is when you actually perform the 
>>> zfs snapshot (yeah, that's implementation specific but you get the idea) 
>>> You two seem to consider it the whole step.
>> Yes, basically you are correct. A checkpoint is a grouping of functionality, 
>> such as "Target Discovery", which is executed as a whole(unless there is a 
>> failure). An analogous checkpoint in DC is the im-pop piece. All IPS 
>> packages are installed at one time, and as far as I know you cannot stop in 
>> the middle of this finalizer and restart.
> 
> From a terminology perspective, a checkpoint as anything other than a point in
> time is confusing to me. A grouping that ends with a checkpoint should be
> called something else, I think. 

Yes, perhaps a different term would help. I shared the same confusion as Jean. 
To me, a checkpoint felt analogous to a breakpoint in a debugger.

Sue

Reply via email to