A uDAPL consumer shall be able to wait on a CNO. Thread-safe providers must allow multiple consumers to wait upon the same CNO. When notified a CNO shall...
834
Paul Hurford
paulhu@...
Mar 27, 2002 2:30 am
Arkady: Is there absolutely no control over Microsoft? How (dumb question notice in advance) can they always get away with this type of extortion? Paul...
Per its bylaws, the IBTA does not define, recommend, or endorse APIs, so the IBTA SC decided this type of session was inappropriate at their event. It is...
My understanding that dual-port HCA will be used for multipathing and automatic failover. As far as Consumer is concern it sees 1 IA. The second port of a...
839
dat-discussions@yahoo...
Mar 27, 2002 3:48 pm
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
Arkady, If the consumer sees only one IA for a multi-ported IA, then there is no problem from a region sharing standpoint. But in general mirroring is ...
... Let me clarify: Register an extremely large LMR (region), clone it for each Protection Zone (PD), and then create RMRs (Windows) with each cloned LMR...
... A good summation. However, there is no doubt that the interception *is* allowed under the API. When dealing with each Interface Adapter the consumer ...
The draft API would be as follows: typedef struct dat_os_wait_proxy_agent { void *instance_data; void (*kick_func)(void *instance_data,DAT_EVD_HANDLE); }...
... Well, let's be careful here. If the provider is not a strict pass-thru, that has both performance and security consequences. We can alter the presumed...
Since "port qualifier" may cause confusion, how about a more generic name to indicate this is really an implementation specific opaque object, such as ...
Is there requirement for the ability to do an infinite wait? If not, should this be added? (There doesn't appear to be the capability to perform an infinite...
... That's not bad. ... I had nearly convinced myself that it was purely diagnostic data, and should be removed. But then I thought about some of the more...
... So then should this agree with the IETF's addressing conventions for IPoIB (http://www.ietf.org/ids.by.wg/ipoib.html)? Presumably a DAT peer on iWarp ...
Ellen is right. While we do support it in APIs we need to add an explicit requirement for it. I will add it for the next version of doc. We can argue about it...
Dear DAT Collaborative members, I would like to strongly encourage DAT members to participate in the ICSC ITWG. This working group defines user APIs. They are...
I have a question about the Send and Receive DTO flag definitions. In Post Send, one of the Completion Suppression flag and Unsignaled completion flags seems...
... Unsignalled completion suppresses the notifiability of the DTO completion, not whether the Event exists or not. Completion Suppression actually swallows...
... From: Deleganes, Ellen M [mailto:ellen.m.deleganes@...] Sent: Friday, March 29, 2002 11:43 AM To: 'dat-discussions@yahoogroups.com' Subject: RE:...
... To be more precise, completion suppression is "don't create completion ***when the operation was successful*** and obviously don't signal when you do not...
Ok, I see the distinction now, but I have still another question. In the case of Unsignaled completion, does this apply to both successful and unsuccessful...
Please find the Infiniband Access API specifications below. This specification combines our OS API, Management Interfaces and CM API into one document. A press...
