( Refer to detailed requirement spec. at http://groups.yahoo.com/group/dat- discussions/files/kDAPL_detailed_reqs_013102.p\ df) 1. Is DAT the only owner of...
Som, We never assumed that DAT is the sole owner of Interface Adaptor. That is why we open and close IA. We do not restrict it nor we require it. We also...
... We could consider an additional requirement along the following lines: kDAPL shall support query methods to obtain native transport handles that correspond...
There not always one to one correspondence between DAT handles and transport object. Lets look at the IB example. What user may really want to know how many...
Exactly ... that is why, for callback notification, the context switch overhead is listed as a 'con'. It is not necessarily a queuing mechanism, but that is...
... There are several scenarios where the Event Queue is far simpler. One is the "user mode only" Event Queue. The NIC itself places the events into User...
I was wondering if interoperability between following items are taken care in DAT requirements, but more I look it seems to me those are details of the layer...
Dear Collaborative, One of the perspective members raised a concern on the copyright coverage of the contributions made by the DAT Collaborative members to the...
Som, Regarding your two points: 1. DAT is transport-independent, so the mapping to InfiniBand is handled by the DAT provider. The DAT consumer has no...
Al, You have provided some useful information..see more comments below! ... Agreed. Question is how? Will there be a different DAT provider implementation for...
Som, My understanding is that the DAT Collaborative will select an appropriate channel abstraction API for its reference implementation, so that there need...
Ladies and Gentlemen, Enclosed is an updated version of the kDAPL detailed requirements to the one I had posted for the ratification last week. Most of the...
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 :...
dat-discussions@yahoo...
Feb 6, 2002 5:32 am
609
This version includes the last posted version of the kDAPL detailed requirements and many changes in kDAPL APIs. The API changes are still not complete. The...
Hello Everybody, now some people will shoot me but: lets take a step back and look into the purpose we want to use DAT for in the first place. Highspeed...
bbialek@...
Feb 6, 2002 12:51 pm
611
Boris' points have expeditious merit, but in my view InfiniBand Architecture's strategic advantage is that it is architected as *a converged fabric*. Unlike...
DAT APIs already provide virtualization. That is why we defined so carefully DAT Consumer and DAT Provider interaction. We explicitly support not only multiple...
Arkady, Agree we should move on past this discussion, but one final point. It may seem shocking, but not all ULPs are so advanced and enlightened as DAT...
5.3 1e is a little strong. Its suggests that concurrent Remote and local reads of the same memory are undefined (proposed new part in quotes). I suggest we...
Rimmer, Todd
TRimmer@...
Feb 6, 2002 5:35 pm
615
5.3 11a I suggest we change "kDAPL 1.0 Provider shall document if it does not support Solicited Wait" to "kDAPL 1.0 Provider shall REPORT if it does not...
I am very happy to report that Banderacom in now an official member of the DAT Collaborative. Welcome aboard. Arkady Kanevsky Chair of DAT Collaborative Arkady...
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 :...
dat-discussions@yahoo...
Feb 7, 2002 7:46 pm
619
The ratified kDAPL requirements document has been uploaded to the DAT reflector. The amendments agreed to at 2/06/02 meeting are indicated by the change bars. ...
My comments on the Endpoint parameters: Protection Zone: This is the only parameter which a) is IN only and b) cannot be known at compile time. Making it a...
We have an action item from the last meeting. The issue is LMR (handle) use for local data buffer specification. Currently we specify LMR (handle) for each...
There is another issue that we discussed before and left for later. Should we specify certain parameters of the endpoint at connection request time vs. at...
Here is my proposed wording for the uDAPL requirement(s) regarding consumer failure. Fire away... :-) Ken - When a consumer process crashes, it is recommended...
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.
