Date: March 10, 1989 To: CAM Committee Participants From: Jim McGrath (Quantum) Subject: Minutes of the Working Group on IBM AT Bus Hard Disk Controller Register Level Compatible Standard Effort - 3/9/89 Throughout these minutes the term "AT controller" will be used when "IBM AT Bus Hard Disk Controller Register Set" (a.k.a. similar to the WD1003 controller chip) is meant. A meeting was held on March 9, 1989, beginning at 9:00 a.m. and ending at 1:00 p.m., at 1820 McCarthy Blvd, Milpitas, CA. Quantum Corporation was the host. The following attended: COMPANY NAME PHONE Adaptec Bob Snively 408-945-6789 Applitech Barton L Phillips 818-341-5595 DPT Tom Treadway 409-830-5522 ENDL Dal Allan 408-867-6630 Imprimis Gene Milligan 405-324-3070 Interactive Systems Todd Reichert 213-453-8649 Maxtor Larry Lamers 408-432-4341 NCR David Skinner 719-596-5612 NCR John Lohmeyer 313-636-8700 Olivetti Dan Crocker 408-996-3867 Optotech Todd Putnam 719-570-7500 Quantum Jim McGrath 408-432-1100 Quantum Art Sklaroff 408-432-1100 Seagate Richard Hoehale 408-439-2619 SMS/OMTI Bob Reid 415-962-2772 Western Digital David Marshall 408-970-9370 The content of the memo calling the meeting was reviewed. Gene Milligan pointed out that there is some standardization activity being done by Conner and Miniscribe in the area of mechanical and electrical characteristics of the AT controller interface (with specific application to embedded AT controller interface disk drives). Gene could not comment on the degree of progress, but did take an action item to get in touch with those involved and report back to the committee. The charge of the committee was amended to include discussions on these mechanical and electrical issues. Gene also agreed to solicit AT controller interface documents from all interested parties (Quantum started off by giving him our OEM manual). This will allow us to cross check any draft standard against current implementations for compatibility. At this point Bob Snively raised a general concern over standardizing on the AT controller interface for AT bus products. His concern was that this particular implementation would not be able to easily support high performance features desired by SCSI host bus adapters. He recommended that we consider standardizing on a new register set for AT bus products. Two points were made in regard to Bob's concern. First, most companies supplying a different interface for AT bus products are not compatible with one another. Thus the committee would either have to include all of these variants as options (thus making the standard meaningless) or break some silicon. The latter is quite likely to ignite a confrontation similar to that we faced in the SCSI committee over connectors (since real, immediate revenues will be at stake). Second, the market is strongly enforcing a requirement of AT controller interface compatibility (over competing hardware level standards). Dan Crocker, as the sole representative of the systems companies, spoke to this point. Some standard will be achieved - the only question is whether the process will be a quick and orderly one via CAM or a long and disorderly one via the market. Nevertheless, Bob had a point. While agreement on any specific hardware level specification other than the AT controller interface is unlikely, it is quite possible that multiple specifications (what Dal refers to as "applications") can be defined at the hardware interface level. This is similar to what we are doing with the OSD interface level, where there are separate specifications for OS/2, DOS, and UNIX. Thus it is quite possible that other committees may be formed to develop specifications for other chips sets for AT bus products, as well as specifications for other busses (e.g. VME). These additional efforts should be viewed as complementing, not competing with, the efforts of this committee. And the efforts of this committee should not be defocused to cover topics more properly covered by other committees. With some agreement on these points of philosophy, the committee then turned to concrete proposals. Tom Treadway had a detailed proposal all ready for the committee, but it had been greatly modified as a result of last minute discussions. Tom started off by diagramming the timing relationships for the current AT bus READ, WRITE, and NO DATA classes of commands. A major criteria of the committee is to maintain backward compatibility for software currently using these commands. By using some of the reserved op codes in the AT bus controller standard, a "super" mode can be activated. In this mode the SCSI functionality can be accessed by passing nexus, command, and status information through the data register using the normal READ/WRITE sequences. The abort command capability of the interface allows data transfers of non-multiples of 512 bytes (e.g. 8 or 32 bytes) for reduced overhead. After going through the sequences for these new AT bus interface commands, the committee agreed that it was a viable architecture. Tom accepted an action item to draft a working document explaining the protocol in detail. This document will be mailed with the CAM mailing and reviewed at the next committee meeting. Major topics which will remain deliberately not addressed by Tom include having the host work in target mode; first party DMA; mechanical and electrical specifications; and an extensive AT bus compatibility check. These items will be addressed by the committee as its work progresses. The next meeting will be hosted by Quantum. Location is 1804 McCarthy Blvd, Milpitas, Ca. Date is Thursday, March 30, 1989. Time is 9:00 a.m. to 5:00 p.m. Suggested hotel (within walking distance) is the Sheraton Silicon Valley East, 408-943-0600. Corporate room rate should be $88.00. See the last CAM meeting announcement for transportation details. Quantum will not be booking rooms, so please make your reservations early.