The ia_close does not explicitly say what happens for asynchronous EVD for ia_handle. I would like to clarify that. If asynch EVD was created as part of the of...
Add a note that RMR_free is allowed for either bound or unbound RMR. If RMR was bound then rmr_free will unbind (free HCA TPT and other resources) and free...
DAT 1.0 was drafted to allow both Implicit and Explicit endpoint creation for the passive side. Under the Explicit model, the Consumer receives a Connection ...
Steve, It does fully capture the proposed functionality. I like the fact that SEVD now does not have DIRECT flag. I suggest that we reverse the meaning of the...
Upon further thinking we unfortunately need one more operation. Just adding Disconnect/Error state is not enough. First, we need to allow Consumers to post DTO...
An alternative to consider, noting the difference between the application requesting the disconnection and the other side of the application requesting the...
Ken, if I understand you correcly you are proposing that the call dat_ep_disconnect when EP got disconected already for whatever reason will transition EP into...
To clarify things, I tried to draw a state diagram (just for disconnecting) for what I am proposing, which lead to refinements to what I was proposing. See the...
Joe, I have posted the updated EVD man pages that included Steve suggestions and notification hierarchy text. Can you read them and comment on them? Thanks, ...
The attached jpg represents my understanding of the changes proposed to the EP (Local Endpoint) object. The new superstate for "Disconnected" can be found in...
Caitlin, 3 questions. 1. Why do we need 3 disconnect states. I am thinking that a single state will do. Drained vs. non-drained is purely implementation...
... The two are related. Assume that draining takes time, and that a reset call is made before draining has completed. If the call synchronously blocks, then...
Just to get proper footing: 1. it_accept is equivalent to dat_cr_accept 2. IT API do not have timeout parameter for ep_connect (or accept) 3. The non-peer...
#include<WINSOCK2.H> is not needed Steve already removed it from reference implementation headers on SF for latest alpha4 drop. I have removed it for DAPL-1.1 ...
The goal is to avoid race condition between evd_modify_cno and event being queued (for example by evd_post_se). Proposal: Only allow to call dat_evd_modify_cno...
... disabled. ... It would be adequate to require the Provider to update the field in an atomic way, hence either the prior CNO or the new CNO gets the notice...
... After an RMR is freed, the IA MUST reject any incoming packet with that RMR's Context. On local operations, however, I believe this must be explicitly...
We should give warning to the Consumers that in-progress posted operations may complete successfully or in error or flushed, and all outstanding posted...
Ken: You're such a rebel. Do you have a leather jacket? Tom Tucker VP, Engineering Ammasso, Inc. 781-982-1702 ... From: Ken Walker [mailto:kwalker@...] ...
Tom Tucker
tom@...
Jan 17, 2003 7:51 pm
1699
That would work. Caitlin, would you please, write an advice to users to be included for dat_evd_modify_cno for it? Thanks, Arkady Arkady Kanevsky...
... How's this? Upon completion of the dat_evd_modify_cno call, the EVD's designated CNO will be altered in an atomic operation. Prior to that instant, any...
We discovered one more problem in the DAT connection set up. DAT specifies that when remote side rejects or when timeout expires then EP transitions back into...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
