> >Copy it, then retry if needed.  The reader needn't serialize. (Gilmartin)
> Sorry, this is not correct. (Relson)

I've been writing OS Assembler since 1967 and have developed a great deal of 
complex multi-task and multi-address space product code and I've always found 
this kind of thing quite challenging and full of unexpected pitfalls. Of 
course, maybe I'm not so smart, but perhaps such matters require more than a 
little thought before asserting how simple they are, especially since to a 
recipient without great experience it could otherwise be quite misleading.


-----Ursprüngliche Nachricht-----
Von: IBM Mainframe Assembler List [mailto:[email protected]] Im 
Auftrag von Peter Relson
Gesendet: Dienstag, 25. September 2012 13:32
An: [email protected]
Betreff: Re: The Transaction state (was Model 2827 New Instructions

>Copy it, then retry if needed.  The reader needn't serialize.

Sorry, this is not correct.

You can only "copy" (and/or avoid serialization) if you know that the element 
you are copying cannot be freed from underneath you (and you can't handle that 
in recovery because you have no way of knowing if your reference will blow up, 
let alone if someone else might have re-allocated the storage and is using it 
for some other purpose that you might or might not be able to detect).

In the general case, readers need to serialize. Not against other readers, but 
against a writer.

Peter Relson
z/OS Core Technology Design

Reply via email to