Martin,

Great explanation and thank you for the idea/ enlightenment. I want to do this 
hopefully once.
I inherited the code and thought I would create a more robust design.

Again, thank you, it's much appreciated

Scott ford
www.identityforge.com

On Jun 27, 2012, at 1:40 AM, Martin Truebner <[email protected]> wrote:

> Scott,
>
> Before I answer your questions:
>
> Why use dataspaces? You need caution when using it (commas!) and they
> are limited (althru way high= 2GB) - New stuff is better off using
> large memory objects (IARV64).
>
>>> And the job ends, does this mean the dataspace created is deleted ?
>
> When the job that created the dataspace ends (regardless how
> privileged it was) the dataspace is deleted. There are tricks to avoid
> that, but a started task (creating it) is much better.
>
>>> Then a STC would read the dataspace to retrieve the stored data.
>
> When the dataspace is created with the right SCOPE it can be reached
> from others. To access it others need the STOKEN for an ALESERV(ADD) and
> can then use the received ALET.
>
>>> What I want is the dataspace created and the job to end ....
>
> and that is the point when its life ends.
>
>>> ...and another process populate the dataspace with data.
>
> The same applies as for the reader of the dataspace.
>
> Something that will work is to have a started task that does the
> create. After creating it publishes the STOKEN. Then it goes to
> sleep and waits for a signal to either
> - do some work ("....retrieve the stored data") or
> - end its life.
>
> Others can use the published STOKEN and place work in the dataspace
> ("...another process populate the dataspace with data") and then signal
> work to the waiting started task.
>
>>> Presently we use another mechanism, basically a subpool to store data,
> but I want to get away from that design and feel the dataspace is better
> suited for the data and the amount of data.
>
> You get pretty isolated storage (and extendadble) with
> dataspaces as well as with IARV64. With either one running over the
> defined boundries will get some kind of program exception (0002, 0004,
> 0010, 0011).
>
> --
> Martin
>
> Pi_cap_CPU - all you ever need around MWLC/SCRT/CMT in z/VSE
> more at http://www.picapcpu.de

Reply via email to