[Info-vax] DLT drives and antique SCSI compatibility (was: Re: Next release of OpenVMS x86_)

[Terry Kennedy]

On Saturday, July 11, 2020 at 1:28:03 PM UTC-4, Stephen Hoffman wrote:
> On 2020-07-10 21:58:28 +0000, Terry Kennedy said:
> 
> > To comment on another reply, I was told by a reliable source that the 
> > TZ30 was the result of a barroom bet between a DECie and a Mitsubishi 
> > engineer about the impossibility of making a half-height DLT drive. 
> > Despite the "Z" in the model name, it is *NOT* a SCSI-compatible 
> > device. In fact, it can hang both halves of the only controller it was 
> > every supported on, the RQZX1. Simple user mode I/O is sufficient.
> 
> That sounds... odd...
> 
> RQDX1 was MFM, and AFAIK was never used with a DLT.

Note that I said rqZx1, not rqDx1. The companion board to the 11/93 CPU.
 
> Various SCSI hardware back then was somewhere between funky and 
> mutually-incompatible, too.

Yes. People were believing that there were things like "DEC SCSI", "Sun SCSI", "Apple SCSI" and so forth. Part of that was ambiguous parts in the SCSI specification and part was that each vendor was implementing their own controllers and drives and didn't do any compatibility testing with gear from other vendors. DEC went further and disn't test compatibility with their own gear, other than specific marketing combinations. In the case of the TZ30 and the RQZX1, that testing wasn't finished when the order came from on high to ship it. By that point DEC wasn't even designing the hardware. That particukar combination of boards was from ROI. I tried to fix what I could in the little time I had - the fact that you can understand the RQZX1's menus is my work, for example.

I have worked on SCSI since it was called SASI, and designed / built / sold host adapters for the S-100 bus. Everything ranging from glorified parallel ports to full sync mode DMA. I also had a hand in the Xebec bridge boards (ST506 to SASI) and worked with ADSI on their physical transceiver chips, which you can find on many devices from that era. Emulex was particularly fond of them, as was Wangtek. That at least cleaned up signalling and arbitration on the bus, but the higher level protocol stuff was still a mess for a long time. I think SCSI 3 is where everyone finally got their act together. The wide popularity and interoperability of SAS drives today shows that everything eventually worked out, but it was a rough road, particularly at the beginning.