Associated links on the Wildflower D* machines from Xerox
Al Kossow's web page contains various fragments of interesting Xerox computer lore.
Dave Curbow, one of the people that worked on Star/BWS, has this web site with some interesting information.
See more about Alan Freier's account of the D* Machines here in his interview for the DigiBarn Computer Museum
See Alan Freier's complete reproduction of the PrincOps Mesa instructions here
The Wildflower Web Site
The machine that used to serve this site is a Xerox 8010 Workstation, affectionately known as a Dandelion. The Dandelion was one a D* machines designed by Xerox in the late '70s and early '80s. Other members of the D* family include:
See our pages on the Dolphin here
- Dolphin - also known as the D0, first D* machine to execute a mop-code
- Dorado - also known as the D1, PARC's long time principle research engine
- Dicentra - a diskless, multi-bus machine used to implement a router
- Daybreak - also known as the 6085, the logical successor the Dandelion design
- Daisy - a VLSI D* machine, never manufactured
- a multiprocessor system used in a high end printing system
The Dandelion is the only one of these processors built to Butler Lampson's wildflower design.
Server physical configuration
Dandelions are not known for for their configurability. The particular one serving this web page has the following optional features:
- 1.5 mbytes of memory
- 40 mbytes of hard disk
- large format display
- keyboard and mouse
Features that are also optional (but I don't have) that might have been appropriate for this service are:
- Large(r) disk (80 or 300 mbyte)
- 23 bits of VM (rather than 22)
- Dumb terminal (replacing display, keyboard and mouse)
The standard components of the system include:
- 2901 bit slice soft micro coded processor
- built in Ethernet
- built in floppy disk
- RS232C interface
- LSEP interface
The processor is a 2901 bit slice design with a cycle time ~139 nanoseconds. For the mathematically challenged yet still caring, if that cycle time was slightly faster, say 125 nanoseconds, it would be an 8 MHz machine, somewhat like the first Macintosh.
The Dandelion is also supported by an asynchronous Input/Output Processor (IOP), which is an Intel 6085. The IOP handles the slow devices, such as keyboard, mouse, floppy and RS232C and is responsible for much of the early stage of the boot process.
The Dandelion was designed to compete with the Intel 8086 class machines of the late '70s. Technically, it did that quite well then, and if you can find a running 8086 today, it'll still run circles around it. In some applications it will also compare favorably with a 80186 or even a 80286, but most of that depends on what problem you are trying to solve and what you are measuring. And it is mostly a software issue, which will be describe later.
The Dandelion was the first commercially available system to come with Ethernet as a default. Though it wasn't the machine that was used during the early development of Ethernet (the D0 gets that distinction) it was the first of the Xerox line to have mass produced Ethernet capability.
The floppy disk is an 8", Shugart 850, double density, double sided device. It was used primarily as a date interchange device. Very late in the life of the D* machines a VM file system was actually implemented that would use floppies for swapping, but that was not exploited to any great degree.
The RS232C capability was provided by the IOP using (probably) Zilog chips. On a good clear day, properly configured Dandelions were known to support speeds up the 56 kbit / second.
Software system
The software running on this machine is officially known as the Xerox Development Environment. Most people refer to it as Tajo. This particular version is modified only slightly from the last version of Tajo shipped by Xerox back in '85 or '86.
Tajo is windowed system that is principally driven using context sensitive pop-up menus. There are, of course, extensions to this paradigm to actually make it useful. The main extension (by design) was the existence of tools. It is the tools of Tajo that make Tajo what it is. Tajo provides a sufficiently rich environment to allow several hundred geographically disperse engineers to develop relatively complicated systems.
The operating system kernel of Tajo is Pilot. Pilot was an experiment born at PARC in the mid to late '70s, then migrated to product development. It was the common base of Xerox Star (later to be Basic WorkStation, or BWS, the various network services, and Tajo.
All the above named systems were implemented in Mesa. Mesa is a programming language and a virtual machine definition. The porting of the latter allowed Mesa based systems to run on the variety of processors of the D* series. For most of lifetime of Tajo, the only language available for development was Mesa. All of the previously names systems were completely implemented in Mesa - there was nothing else. Commercial languages, such as C, Basic and Fortran, were developed to run in a Mesa environment, but were never (to my knowledge) used for serious development.
It should be noted that the Mesa virtual machine was implemented in a combination of Mesa and microcode. To keep that in perspective, particularly for the Dandelion, there was four kilowords (40 bit words) of control store. That 4K provided the lowest level interface for the high speed devices (display, disk, ether), as well as the Mesa instruction emulation and some aspects of the Mesa runtime.
Ethernet
Two features were unique about the D* machines at the time: the bit map display and the Ethernet. Both had been employed on the now famous Alto personal computer that was also developed at PARC by Chuck Thacker, et al.
Ethernet was developed at PARC, the dream of Bob Metcalfe, who had previously been toying with the Aloha protocols. Engineering assistance was provided by Dave Boggs, et al. Originally running at 2.94 megabit per second, this version of ethernet was still heavily used up until the Dorados were decommissioned. 10-BASE-5 was developed in the late '70s using the Dolphin as a platform. The Dolphin Ethernet controllers were never produced in quantity, and the Dolphin implementation was rarely employed even internally, and certainly never as a product.
Early in 1980 the collaboration of Digital, Intel and Xerox brought about the public unveiling of Ethernet pretty much as it is today.
Original drawing by Bob Metcalfe
Teaching Tajo
The following links point to actual text of tutorials for new users of Tajo. These are mail messages that were periodically send to the distribution list, "NewUser" to introduce new users to Tajo.
- Booting. It covers a variety of issues on both the 8010 (Dandelion) and the 6085 (Daybreak) processors. It defines the terms unique to the environment and possible configurations that one might encounter.
- Basics of using the Tajo environment, manipulating windows, loading files, etc. It is taught using the mail tool as a vehicle. The Tajo window system isn't like most others.
- File manipulation in Tajo, particularly with the executive. This covers multiple volume, multiple directories per volume, and Tajo's version of a search path.
- Debugger. Tajo's debugger (designed in conjunction with the Mesa compiler and Pilot, is still one of the best debuggers I've ever used.
- Compile Bind Run for Xerox 8010 Dandelion
See Also:Back to Wildflower site
Dave Curbow's Xerox Star Historical Documents
The Digibarn's page on the Xerox 8010 Star Information System
The Digibarn's extensive collection of Xerox computers and other artifacts
Please
send site comments to
our Webmaster.
Please see our notices
about the content of this site and its usage.
(cc) 1998- Digibarn Computer Museum,
some rights reserved under this Creative Commons license.