All, I'd like to follow-up on our discussion in this week's teleconference about adding support for stacked-die memory devices within the existing IBIS [Component] section. While it seems simple for a new keyword to indicate multiple instances of a particular IBIS [Model] behind a component pin, the main consequences for EDA tools have nothing to do with parsing such a keyword or netlisting extra I/O buffer instances to a simulator. The issue is that much EDA code, both ours and doubtless elsewhere in the industry, has been written under the assumption that one I/O buffer lies behind an IBIS component pin. User interface dialogs for assigning models to component pins assume this, code to identify driver and receiver in automated batch analysis of boards assumes this, code to save and retrieve I/O buffer assignments and direction settings assumes this, etc., etc. It's not that our developers cannot cope with requirements to make database and UI changes in these areas, but that we need to think carefully about the payoff versus engineering investment. I'm afraid that the payoff from tracking down and making the many tool changes for this will be very limited when we already commit ourselves and our developers to making EBD a more general interconnect format to handle multiple dice within a package. The overhead to IC vendors of shipping an .IBS file with "bare-die" component together with an enhanced .EBD file for a stacked-die device doesn't seem too bad given that they must bundle at least one extra file containing ISS subcircuits in any event. So I would in particular ask the other EDA vendors on the committee to think about this trade-off and let us know your perspectives. Thanks, John