NT

The final selection of NT as a platform to host the ZIY has many practical and historical reasons. When I originally started work with the Metrology Group, we were attempting a proof of concept device, with little thought of interfacing it with anything. Why, after all, work on interfacing something that doesn't work? Our mission was to first prove that an open-air laser based metrology system had some chance of working. We thus concentrated on quickly and inexpensively putting together a prototype instrument using COTS (Commercial Off-The-Shelf) technology [4]. Using a PC running DOS was a logical extension to this approach. PCs then were relatively simple and inexpensive, and the programming tools then available for DOS were excellent. As it became clear that the laser metrology concept stood some chance of success, emphasis was shifted to designing and refining further prototype instruments, software and techniques. When we added more than one instrument to our experiments, it became clear that the software would have to be split into what would become the ZY and ZIY modules. PCs running DOS were quite adequate and convenient for the ZY software, and communication between ZY and ZIY was over an RS-422 serial link. To this day, the ZY software is a DOS application, and though Ethernet has replaced the RS-422 link, this arrangement has been quite satisfactory. For the ZIY, however, DOS proved to be a rather limiting environment. The precursor to the ZIY provided an interface to multiple ZYs while providing a user interface, and it became awkward to program this in the single tasking DOS environment. At the time we had few alternatives for the environment that would host the ZIY:

• VxWorks, on a Motorola 68000 based VME system
• QNX, on the PC architecture we were currently using
• Unix

All of these options involved a steep learning curve for members in our group. At the time we were concentrating on the instrument rather than a metrology system as a whole, so we lived with the limited DOS environment and put off dealing with this problem. The ZIY software was strictly throw-away code. This throw-away ZIY eventually evolved into a Windows 3.1 program that served the metrology group well. At one point we did attempt a permanent ZIY hosted on a Unix system, but for a variety of reasons this did not work out well, and before we could work out the difficulties, we lost our Unix programmer. We were back to the throw-away ZIY. By this time Microsoft released Windows NT 3.1. It was touted as a solid multitasking, multithreaded OS. It included TCP/IP networking out of the box. It provided a familiar GUI. The Win32 Application Programming Interface (API) was familiar, in that it was based on the Windows 3.11 API that I already knew. It ran quite well on a 486 DX2/DX4 based PC, and was inexpensive, relative to other x86 solutions. It seemed perfect for the job, at least on an interim basis. As it turns out, it proved to be as dependable and flexible as it promised to be. Though we were still focused on the laser ranger instrument, in November of '95 I undertook to develop a brand new ZIY for Windows NT, with the intention that this would not be a throw-away version.