Minutes: IEEE P1285/X3T9.2 DADI Grass Roots Meeting July 13, 1992, Santa Clara University, 1:00-3:00 PM The meeting participants were: Jim McGrath, Quantum Tom Hanan, Western Digital Shishir Shah, Western Digital Larry Lamers, Maxtor Chris Hamlin, Apple Andy Hospodor, Santa Clara University Martin Freeman, Philips Research Dave James, Apple Dennis Pak, Apple Scott Smyers, Apple Mike Teener, Apple Martin Freeman convened the meeting and mentioned guidelines for the meeting that were suggested by Don Loughry. He also mentioned that if more "Grass Roots" meetings were desirable, then that could be arranged. The following seven discussion points were mentioned by the participants: * What is P1212/P1394 DADI Interface * What is P1285 current status * How do both conflict with the Loughry recommendation * How do we work together (press releases) * Liaisons, Co-Located Meeting (Nov?), FAX, Email * PCMCIA and "Card Bus" * Dividing on logical level (?) (logical-electromechanical) Tom Hanan suggested that we prioritize these discussion points to make sure we covered the most important topics in the present meeting. The following three discussion points were chosen: * What is P1212/P1394 DADI Interface * What is P1285 current status * Liaisons, Co-Located Meeting (Nov?), FAX, Email P1285 Status: ------------ Martin Freeman and Dave James talked about P1285 including answering questions in an interactive discussion. P1285 currently has the following characteristics, but is not limited to these characteristics: - Command and user data space is memory mapped - One controller, multiple slaves (today) - Memory mapped disk - Caching control - Performance Goal: 16 drives with 25 MB/sec per drive (peak) - Main memory and secondary storage merged - Addresses may be treated as logical or physical (not decided yet) In the area of implementation levels, the current thinking is: High Priority: System Software Low Level Protocols Medium Priority: Electrical Low Priority: Mechanical Jim McGrath inquired about the possibility of the mechanical level for P1285 being done in the Small Form-Factor Committee. There was a general feeling that this would be one good way for P1285 and X3T9.2 to work together, where the work would be complementary. P1212/P1394 DADI Interface: -------------------------- Scott Smyers, Mike Teener and Jim McGrath all made contributions to this presentation. Presently, there are three related efforts: DADI, serial bus protocol (SBP), and SCSI on Fiber Channel (GPP). The common thread is that they are all concentrating on serial protocols. However, DADI has a larger scope which presently includes the initial concept of DADI which is similar to P1285 and its new emphasis on P1212/P1394. X3T9.2 people like the concept of P1394 but don't like the P1212 interface. They have a software investment in another interface. P1394 is seen as a transport mechanism by which the SCSI protocol can be transported to a device. Basically, the DADI committee is thinking of working on a new DMA structure for P1394. It looks like the memory and protocol interface typically found on a LAN controller. The host prepares a linked list of commands and data, and tells a target device where to find this linked list. The target by its own means fetches the linked list from the hosts address space and performs the indicated operations. As these commands are executed the target constructs a list of responses, and the address of this response list is passed back to the host. The DMA action involved in this is in passing the control information back and forth between the host and the target device. How the data is moved or manipulated is not part of this interface. It is a command and response interchange specification where information is passed by reference. There was some discussion over the term "memory mapped" used by both P1285 and X3T9.2/DADI however differently. P1285 uses the classical definition of mapping I/O units into processor memory space. X3T9.2 uses "memory mapped" to mean that command blocks in main memory have their addresses passed to I/O units. The SBP is another effort within X3T9.2. It is a similar serial bus for transporting SCSI commands, but not quite the same as the proposed P1394 DMA control structure. Right now X3T9.2 is trying to decide whether to merge these two effort or not. It is also trying to decide whether the P1212/P1394 DADI effort really should be handled directly by X3T9.2 or by a new P1394 Industry Trade Association that is about to be formed. Thus, we noted that the current DADI interests are in first party DMA while P1285 has not settled on a DMA architecture. We also noted that if the P1394 DADI work becomes associated with this new trade association, then the focus of X3T9.2 might go back to the original DADI concept that is similar to P1285. Both the X3T9.2/DADI and P1285 participants understood the dynamics of the situation, and that since both efforts were in their formative stages and the changes in technical direction might occur in a short time span. The SCSI on Fiber Channel effort was not discussed. Liaisons, Co-located Meetings, FAX, Email ... --------------------------------------------- The grass roots meetings were seen by the participants as a means of discussing procedural matters and ways to facilitate cooperation. These meetings were not seen as a forum to make technical decisions or negotiate turf. In this regard, as both standards efforts are in their intial stages it was decided that for the time being that there should be a "grass roots" meeting on a quarterly basis. Martin Freeman proposed a meeting for October, while Jim McGrath proposed a meeting for September. Tentatively, the meeting was set for a morning during the week of September 28. Both Martin Freeman and Jim McGrath will coordinate with each other and establish the exact date. This next meeting will be hosted by Quantum. A major topic at this meeting will be a co-located meeting between X3T9.2/DADI and IEEE/P1285 this November. Since X3T9.2 will be meeting in Sunnyvale this November, this would be an excellent candidate for a co-located meeting. The notion of a "grass roots" reflector was mentioned as a means for the grass roots committee members to communicate with each other and send minutes. Mike Teener volunteered to set up this reflector. There was a discussion about a P1285 reflector separate from the "grass roots" reflector, and all thought it was a good idea. Martin Freeman will look into arranging for such a reflector. The issue of liaison representatives was discussed. Jim McGrath suggested several groups that P1285 could possibly liaison with. Larry Lamers narrowed these suggestions down to three: X3T9.2, PCMCIA, and P1394. Scott Smyers volunteered to be the P1394 representative to P1285. Martin Freeman will look into establishing liaison representatives with these groups. It was understood that PCMCIA was not an official standards body, and the importance of such a liaison was less. It was noted that Martin Freeman could request such liaisons, but the actual appointments would be made by the respective groups. The participants felt that because there was a on-going "grass roots" meeting activity planned, and since there were to be liaison representatives appointed between X3T9.2/DADI and P1285, that two co-located meetings per year would be sufficient to keep both sides informed. Jim McGrath mentioned that the word "disk" and the word "storage" were sore points with several members of X3T9.2. He suggested that the word "memory" be used instead of the word "storage". He also suggested using the term "high latency" with regard to memory as well as emphasizing the word "non-volatile." The P1285 participants took this suggestion seriously and will look at alternate terminology. Martin Freeman related a recent telephone conversation he had with John Lohmeyer about the same subject. Martin felt that John's position was that the name change was not meaningful unless there was substance behind it, namely that P1285 would not do a disk interface. Others noted that charters are harder to define accurately, since the distinction between bus standards, memory controllers, and storage devices is much less clear now than in the past. Mike Teener and Dave James related the concerns of many large organizations towards particular names and cited examples of standards efforts where the word "network" could not be used, even though the work involved networks. It was recognized that the word "disk" is a similar "high-sensitivity" word within X3T9.2. In a discussion about the scope of P1285 which includes Flash memories, Jim McGrath noted that X3T9 has no historical charter of interest in Flash memories. As mentioned in these minutes several recommendations were made by the participants to both the IEEE/P1285 and X3T9.2/DADI committees: 1. Recommend P1285 use the term "memory" and DADI refrain from using the term "memory" when describing their standards efforts. 2. Recommend a "grass roots" reflector be set up. 3. Recommend "grass roots" meetings initially to be held four times a year and co-located P1285/DADI meetings held twice a year. 4. Recommend the first co-located meeting to be held during the November X3T9.2 meeting week in Sunnyvale. 5. Recommend the next "grass roots" meeting be held during the week of September 28. 6. Recommend liaison representatives between P1285 and X3T9.2, PCMCIA, and P1394. Martin Freeman