Hank.Walker@CMU-CS-VLSI (02/04/83)
I have recently been to several meetings on topics that might interest you or others. The first talk was by Jim Morris, of the Information Technology Center (ITC), the joint CMU-IBM center for developing personal computers on campus. They recently knocked down an old building and are putting up a new one for the ITC. The people involved from IBM appear to be quite sharp. Morris is from Xerox PARC, and is the head of the center for CMU. A friend of mine Jim Gosling recently went to work for IBM at the ITC. He is the author of Gosling's EMACS, a better editor you won't find on ASCII terminals. He is one of the best hackers that I know. The other IBM personal come with experience in networks and the IBM PC. IBM is serious about this. The basic plan is to spend most of the manpower developing a network system. The system hardware will be IBM rings on a backbone network. Print servers, file servers, and mainframes will hang off of these along with the PCs. The PC hardware isn't definite yet. The interim machine is the CS9000, a 68000-based machine. The network software will use some version of SNA. The idea is for the network to appear invisible to the local machine. Files, printers, and computing resources reside transparently on it. A local disk may be available for caching. The rest of the software will mainly borrow from existing stuff for now. The message passing stuff from the Accent operation system of the SPICE project, EMACS for an editor, Canvas from SPICE for a windowing package, rdmail or hg-like mail system, etc. The schedule is to have several hundred interim machines with final machines to start arriving in 1985, with 1000 installed the first year and ramp up to 8000 quickly. Most of the IBM people are in place, CMU people still being hired. Building done in July. The second talk that I went to was by Steve Jobs and Larry Tesler of Apple on the Lisa computer. I was quite favorably impressed, much more so than I had expected. Apple has sunk 50 megabucks into quite a nice machine. The basic hardware is a standard keyboard, 1-button mouse (of their own design), 6MHz 68000 CPU, 1Mbyte memory with parity (no ECC), 5Mbyte Winchester, and dual 860K minifloppies. It also has little things like a clock/calendar, speaker with tone generator, and dual serial ports. The display is BW 364 lines by 720 dots on a 12 inch diagonal screen. Refresh rate of 60Hz. The screen was surprisingly sharp considering it is only 1/3 the resolution of a PERQ screen. They implement virtual memory by having procedures go through a jump table and trap if non-resident. Data must be resident and is slow swapping in. User has control over this. Virtual memory size is now 16Mbytes. The performance limit is the Winchester. Graphics for text was a little slow, and sometimes pretty slow for figures. It is done in assembly code while the rest in Pascal. No special graphics hardware. The software is the impressive part. The demo that they gave, and the brochure emphasize the office software. A visicalc, graph drawing package, word processor, drawing package, PERT chart system, and mailing list system form the basis of this. All run through a standard interface with windowing and mouse usage. Basically a new and improved version of the software you find on Altos and the Xerox Star. A development environment for Pascal/Cobol/Basic/Smalltalk is available with its own editor, debugger, etc. Also Xenix and CPM environments and a terminal emulator (VT52/VT100/3270). A new language Clascal will come out. It is Pascal with Classes a la Simula. Random information: Smalltalk is very slow. Only one button on the mouse makes it easy to learn but sometimes need too hands. The mouse design seems more relaible than the Xerox mice, but not as good as an optical mouse. The keyboard is thick and key feel is crummy. There is no key labeled as the control key, but the shift lock key can be used as one. In the future they will have a 1 megabit/sec Ethernet-like network with a node cost below $500. They will then provide mail, print servers, etc. Currently there is little software locking for multi-tasking. Instead the programmer inserts explicit statements when he reaches an idle state (waiting for input) so that another process can take over). This will change in the future. About half of the Winchester is taken by the system software. No color graphics this year. Backplane specs and OS interface specs will be published to encourage 3rd parties. Not all of OS will be visible. They don't have floating-point. They plan on sticking with future generations of the 68000 and whatever peripherals Motorola puts out. They will start implementing on-site field service as well as the existing carry-in service. Has built-in diagnostics and diagnostics floppy. Easily disassembled hardware. No remote diagnosis. The dot matrix horizontal resolution is 1/2 that of Dover/9700. The list price is $10000 plus software costs. Apple's emphasis is on the office and making the system usable within 30 minutes for a novice user. I did notice that the marketing guy demoing it did have some difficulties selecting windows and occasionally needed pointers from the software guy. The whole project ran around 3.5 years and 250 man-years for software. One unique feature of the document processing is that documents have an intended target printing device. If you want to print on another device, Lisa will do the best it can, but will warn if something is bogus (italics on a Daisywheel lacking them). Steve Director talked to Jobs for a while this morning, so I'll find out if there is any more dope. I think that there basic purpose was to try to get CMU to take some Lisas as interim ITC machines and see that there was a growth path to nice machines in the future. Considering that they sent out a whole boatload of people along with the company president and several machines and fancy demo means that they consider it important.