Roland, > From: Roland Dreier [mailto:[EMAIL PROTECTED] > > If this were some feature that allowed us to do something new, or made > applications more efficient, or something like that, I'd be all for > it, specs be damned. But in this case it's just bloating driver code > to work around buggy applications. And I'd rather use my I$ for > something more useful. (And in fact the proposed change is itself > buggy -- it calls any completion on the send queue a send, even if it > was actually something else like RDMA read/write, atomic, etc) >
>From a pure technical point of view, I agree 100% with all your comments. In fact I had alluded to all these issues in the original 1 line submission. However, I think you are missing the point. <soapbox> In order for Infiniband to be successful, it must expand the set of applications which it can run. This often means making technical compromises to permit expanded market share. In the case in point, I'm working with a 3rd party whose application already was ported to VAPI, the SilverStorm stack, and will work on OFED/ipath. In the interest of cooperation, I'm trying to help so that OFED/mthca can also run this application. Telling the customer they must redesign the application for OFED/mthca, is not practical and certainly does not promote a positive view of Infiniband for the market. Yes the application is not perfect, yes it should be rewritten, but the customer can just as easily move to other alternatives. We need to promote IB in the marketplace, not admonish customers who are trying to use it. It is this lack of portability of applications which has significantly hurt infiniband's reputation in the market. While its nice to focus on building the perfect implementation, sometimes compromises are necessary to ensure the overall success of the technology. VHS vs Beta is a perfect example. </soapbox> Todd Rimmer Chief Architect QLogic System Interconnect Group Voice: 610-233-4852 Fax: 610-233-4777 [EMAIL PROTECTED] www.QLogic.com _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
