... If Consumer code have to account for this behavior (earlier unblocking before timeout expire) then they can not rely on threshold value other than 1. We...
... Good suggetion. Making it a provider attribute will discourage claiming "I can't" when in fact they should be saying "I'm lazy". So the proposal would be...
I think that Provider attribute for the maximum threshold value is can support would still work. 1. Providers that are not on HW that support queue depth query...
I would like to welcome the newest member of the DAT Collaborative - Ammasso, Inc. Welcome on board of DAT train. Arkady Kanevsky, Chair of DAT Collaborative ...
During this weeks conference call it was agreed to explore the possibility of a polymorphic interface between the low level EVD (S-EVD) and the high level EVD...
I also thought that we agreed to it long ago and reported that to DAT. several of previous emails from Caitlin, me and others have specificially talked about...
Mark, Most Consumers also want to know if the library (IA) they open thread safe or not before they open it. Another words, you do not have a handle yet. In...
DAT seems to be very flexible in that regard, and I'm just trying to understand why ITWG would wish not to offer library choice flexibility. Isn't "library...
Mike, API support for both thread-safe and non-thread-safe implementations has been agreed to since the early stages of the requirements phases (refer to ITWG...
Hi Mike, I could live with that. But the compromise had been to allow the vendor to supply either or both flavors of implementation. What I really want to...
Peter Hochschild
phoch@...
Oct 4, 2002 10:51 pm
1524
Would it be too egregious to specify a naming convention for IA names -- similar to the way it is done today for thread-safe vs. non-thread-safe shared...
Tom Tucker
tom@...
Oct 5, 2002 3:18 pm
1525
Hi Tom, To my mind, that isn't attractive. Actually your note reminds me of another set of reasons I dislike the mixed-safety model: dragging adapter ...
Peter Hochschild
phoch@...
Oct 6, 2002 5:48 am
1526
Peter: I admit that a naming convention for IAs that reflects the thread safety of the provider library may be inelegant. It is clearly a hack to remove the...
Tom, In my view, the current stake in the ground is that the consumer can query the API as to its thread safety. I think the query would be used (if at all)...
Peter Hochschild
phoch@...
Oct 6, 2002 7:05 pm
1529
Peter: Ah. I thought the goal was to specify a portable method for applications to *choose* the appropriate provider library. I believe that DAPL should either...
Tom Tucker
tom@...
Oct 7, 2002 12:19 pm
1530
Steve: I think it's obvious that the application writer makes the thread-safe vs. non-thread-safe decision at design time. After all CODE needs to get written...
Tom Tucker
tom@...
Oct 7, 2002 2:22 pm
1531
Steve: Cool, I see how that works -- It also answers Peter's concern (mixed thread safety libraries) by having the libit library load only libraries of one...