[concurrency-interest] Project Amino: Introduce Lock-Free Stack

[Gregg Wonderly]

Jed Wesley-Smith wrote:
> Hi David, my previous email didn't address your points directly:
> 
> David Holmes wrote:
>> But for that matter does a concurrent stack make sense? What defines a
>> stack? The fact that items are removed in the opposite order to which 
>> they
>> are added. But in a concurrent environment how can you tell? :) Maybe 
>> all we
>> really need is an approximation of last-in-first-out, and a sequential 
>> stack
>> is a sufficient but not necessary solution to that problem.
>>
>> Ditto for a ConcurrentQueue or ConcurrentDeque.
>>   
> 
> As I pointed out in my previous reply, they probably only don't make 
> sense if strict fairness is required, besides you particularly need to 
> know if you want to use them with single or multiple readers or writers.
> 
>> At this point I await someone who is really an expert on concurrent data
>> structures to chime in :-)
> 
> Let's hope so!

I think it's probably not a good idea to consider that all methods on a List 
will be used from all types of concurrency.  Rather, I'd like to suggest that 
perhaps a non-concurrent environment might use index methods to sort or 
otherwise rearrange a list contents, and then publish this list into a 
concurrent environment, to be consumed as a "work" queue or some other such 
strickly end-to-end traversal/consumption.

The lifecycle of data structues in a concurrent environment is hardly ever 
single phased.  There are often multiple phases, and each phase can present 
different requirements on the operation of the data structure right?

Gregg Wonderly