[Stackless] Google Application Engine Thread on Stackless

andrewfr_ice at yahoo.com andrewfr_ice at yahoo.com
Fri Mar 27 00:47:25 CET 2009


Hi Rana:

--- On Thu, 3/26/09, Rana Biswas <xdevice at gmail.com> wrote:

RB> Won't saving the data in a database be a
RB> better idea in this scenario.

I suspect a lot of the WS-BPEL processors do this. However most processors are rather big and complicated. However Apache ODE takes the execution persistence approach. Here is a URL:

http://ode.apache.org/jacob.html. The problems are explained (by the way, I spoke to one of the implementors. He was impressed by Stackless).

However lets talk this out. Maybe there is something I have missed?

Pretend you have a simple Stackless programme consisting of two tasklets, "FOO" and BAR. 

def tasklet("Foo"):
     A()
     B()
     C()
     D(<blocked>)
     E(..)
     F()

def tasklet("Bar"):
     A()
     B(<blocked>)
     C()
     D()
     E()
     F()

The tasklets have state (the function are doing a lot of computations). At some point, you have decided to serialise the programme. When data is ready  how would you write the code so you can easily resume tasklet Foo from d() and Bar from b()? Perhaps you could write the processor as a state machine but that is cumbersome. Also how does a database help you if the programme has radically altered?

Cheers,
Andrew



      




More information about the Stackless mailing list