|
|
As the company moved from organisation to operation, communication which had previously been conducted through the Information Letters moved more and more to verbal communication at the monthly meetings of participants, as well as ongoing telephone communication among project teams and messages on our teleconferencing system (“MJK”). Dan Drake inaugurated the publication of summaries of the monthly meetings with this document.
This was particularly eventful time. The decision to scrap the QBASIC version of Autodesk and begin rewriting it in PL/I was made. Development of the initial version of AutoCAD was coming to a close; we decided that versions would be developed for the Scion Microangelo, IBM PC, and Victor 9000, and that the initial price would be $1,000.
The August general meeting was held on Sunday, August 1, 1982, at Jack Stuppin's house.
The meeting was called to order at 1:10. All company people in the Bay Area were present: Dan Drake, Mike Ford, Dave Kalish, Duff Kurland, Mauri Laitinen, Greg Lutz, Keith Marcelius, Hal Royaltey, Kern Sibbald, Jack Stuppin, and John Walker.
Stephanie Nydell, whose interest in joining the company was discussed in the previous information letter, was introduced to the meeting. There was unanimous agreement on bringing her in as a founder, subject to approval by the people out of state.
There was a brief discussion of the mechanics of buying back the shares of Jodi Lehman, who has decided to leave the company. It is possible that the company can buy them back, or a new participant could buy them. It appears that there will be no difficulty in working out the details with legal counsel.
There was no dissent from the published minutes of the July meeting. John Walker presented the financial report. The balance of funds last month was $45,592. Expenditures have been as follows:
Supplies 85 Softcard for Apple 382 Digitizer for MicroCad project 665 C compilers 42 Apple documentation ? ------------------------------------------------- Total 1,671 Income: Interest from CPF 325 Current Assets: 44,426
Marinchip Systems, Ltd has for various reasons bought items for ADI
worth about $1,542. ADI has also agreed to buy hardware from Duff
Kurland in the amount of $5,115.![[Footnote]](i/footnote.png){kind=small}
This makes:
Current Liabilities: 6,657
Planned expenditures include compilers (C, PL/I, CB80), business supplies such as invoices, and computer equipment, as yet unspecified, for the use of Mauri Laitinen.
Everyone delivered a report of 2–3 minutes covering:
Progress on various projects will be discussed later. An important feature of the reports was that there were almost no reports of obstacles to getting the work done.
There are some CP/M utilities available to anyone who wants to use them within the company. Some are not ADI property and should not be distributed outside.
We have the current IBM software distribution agreement. Anyone who is curious about their standards can get a copy.
In addition to Mike Ford's connection with Victor, we have good contacts at Apple, Hewlett-Packard, Onyx, Corvus, and Timex (which sells the Sinclair). Most of these are just waiting for us to have products to demonstrate.
Dan Drake brought up a possible guerrilla project: an existing Marinchip program that computes all known tax deadlines. The complexities of tax payments are such that the Wall Street Journal ran a flowchart on Federal withholding tax deposits alone. A program that handles these details ought to be worth $50 to any business in California. (It would be a major project to configure it for other states.)
The consensus was that we should convert the program to CP/M and put it on the Apple Softcard. We should get an opinion from accountants, though they have a conflict of interest in that they get high fees for providing the same information to their customers. The program will have to be sold with the understanding that it will be obsolete in a while; people will have to get updates when the rules change. It would be appropriate to let someone else distribute it, since it's off our main line. A major problem is putting on a good enough disclaimer to keep from being sued into nonexistence if we, the customer, or the tax people misunderstand the rules.
John Walker described Micropro's anti-theft provision: their manuals
say, in places that make the statement hard to get rid of, that if
your copy of Wordstar doesn't have such and such a sticker on the
label, you have an illegal copy. Just send it to us and we'll give
you a legal one! Of course, they can find the buried serial number on
the disk, so they know who the fink is. This seems the best
protection scheme in the business, and well worth emulating.
The most animated discussion of the day was on marketing MicroCad, which is very close to being a reality. The plans agreed on are summarized as part of the MicroCad project discussion.
There is now a functioning calendar, and the data base is being redesigned. After a short discussion of features and problems, John Walker raised what seem to be fundamental problems in the project. What follows is approximately a logical, not a chronological, summary of the discussion.
The original quick and dirty implementation in CBASIC is now running on CP/M systems in CB80, and Kern Sibbald has put hundreds of hours into conversion and making it run well, but we seem no nearer to a salable product than we were at the Computer Faire. It's still of no practical use, being unreliable and terribly slow.
There was disagreement on the extent of the non-progress, but a consensus that things were not moving nearly fast enough.
The version at the Faire looked like an outstanding product, but it glossed over many crucial technical problems which must be resolved before anything is sold. These don't have obvious answers, and the attempt to fit the answers in as we go along has given us a program as big as MicroCad that does less and runs slowly.
One technical problem may be the wrong choice of language. CB80 lacks the right I/O facilities and requires fairly massive assembly language interface routines, which are especially clumsy because of the lack of data structures. In C or PL/I the problem would disappear. But that conversion would take some time.
Perhaps the existence of the prototype has fooled us into thinking that we could work that into a product, when actually a full re-design is needed.
If we scrapped the project entirely, we would have plenty of things to do with the manpower released, but no one wanted to do that. On the other hand, if we continue the project, we need more people involved in it.
A consensus was developed along these lines: The program will be rewritten in a better language. At present, if we want to sell to half million existing 8080 systems, that means PL/I; by the time the program is ready there should be a PL/I available for the 8086. Also, when the screen handler has been done in PL/I, we'll have gone a long way towards a PL/I version of Window.
The thing we are to produce is a user-friendly card box system with multiple boxes holding cards of unlimited size. Once it's done and on the market, we can work on further releases with added features. This first implementation may get us into some decisions that we'll regret when we start adding features, but we have to get a working product out the door in a finite time.
Dave Kalish will work with Kern Sibbald on the database design and on user features. Duff Kurland and Mauri Laitinen will work on the screen handler. As a crash project the thing could be done in a month or two; since we don't have people working full time, we must be resigned to its taking longer.
There was general agreement that we needed to know more about competing products, including MBA. No one actually said that he'd do such an investigation.
We now have data on availability. The price quotes range from zero to $3,450. We are supposedly getting a free copy in 6–8 weeks, but no one is betting on it.
Both the PL/I version on the 8080 and the C version on the IBM PC are now converted and running, needing some amount of work before release. The Victor version will be running soon. Without any optimization or division into overlays, the 8080 version has 4K bytes of memory left over, and runs as fast as the 9900 version.
The discussion of marketing went very roughly as follows.
John Owens has sold perhaps 20 copies of the 9900 version through his ads in Byte, and has had lots of people ask why it isn't on the 8080. He might sell as many as a hundred in a short time if we turn him loose.
There are some large markets for an 8080 MicroCad. Scion is selling
enough Microangelos to support a color ad in the front of Byte every
month, though there's very little software available. A version using
the Microangelo with a light pen would be a completely standard
product that would run entirely on Scion's graphic equipment and might
be something they would want to market. There are other markets
further along, like Apples and the incredibly cheap Sinclair. Again,
it might be easily converted to Univac PL/I, running on an 1100 with
Tektronix graphics.
On the other hand, since we could ship out copies and sell a few almost immediately, there are marketing decisions that must be settled immediately. What do we do about Beta testing? Do we sell a pre-release? How do we price it? What do we know of the reactions of live users of the 9900 version? The only people in the company who have spent any time using it are computer fanatics who know the internals and are not well qualified to judge user interfacing for a large market.
It's clear that getting some copies into the field and getting
information back from end users will be invaluable in developing the
product.
Getting some actual money into the company is
also important.
Fears of getting a hostile review on an early release were dissipated by the observation that it's extremely hard to get reviews when you want them. A product that's selling at most a few hundred copies is in little danger of any reviews.
The consensus reached was this: The 8080 version will get a driver for
whatever plotter John Owens currently likes to sell.
Stephanie Nydell, who has done documentation and support for larger
graphic systems in the electronics business, will write up suggestions
for making the user interface more friendly to completely
unsophisticated users. She will also make minimal changes in the
current manual to make a preliminary version that we can send out with
the early release. We'll tell Owens that the product is coming, and
he can sell it to his 8080 prospects. On Wednesday, August 25, we
want to send this version out.
There will be a Victor 9000 version for Mike Ford to demonstrate by the end of September.
The initial list price will be the same $1,000 as for the current 9900
Interact.
Once we have this product out the door, we demonstrate to Hewlett-Packard and pursue our other contacts. Work will start then on a new manual and will proceed as we exercise the program and get feedback from users.
META is now running on the 8086 but has not been modified to give some features that the QBASIC compiler will need for compatibility with CB80. The corresponding compiler changes have not been coded.
Pass 2, the optimizer, now generates 8086 assembly language. Now that we have the definition of the relocatable format, QP2 must be modified to generate relocatable directly, as well as supporting the new CB80 compatible features.
The memory model is well defined, using the Large Model for 8086 addressing. Parameter passing is mostly defined, but not all decisions have been made.
Queue handling and buffer allocation are working. The floating point library is working after a pretty complete rewrite and will support 32 and 64 bit integers if we want them.
There is a large amount of work needed on the library, without enough people to work on it. This is also the most easily partitioned part of the product.
To get all the parts of the project moving, there was a general reshuffling of manpower. Dan Drake will do the compiler pass 1 and associated changes in Meta. Mike Riddle will rewrite pass 2 in C; optimization will be based on the current QP2, and the output will be relocatable. Greg Lutz will work on parts of the library.
We will have a working version for Victor at the beginning of October. The big question is whether Digital Research will get CB86 out before we get ours out.
The in-core version of line database is running. There is disk code, which is not running yet. There are some problems with string linkage and inadequate documentation of CB80 linkages.
The consensus was that the Z80 version would be ready in a month. The manual update should be an afternoon's work. The 8086 version will be out when QB86 is.
Stock options are being set up with corporate counsel. Most of the paperwork will be ready to execute at the next meeting.
The process of unanimous decisions was discussed briefly. The unanimous agreement procedure that we adopted was intended for the adoption of new participants, not for other corporate decisions that the management or the board can make. If the organization gets large, polls of everybody will get too awkward. As we don't expect to bring dozens of people in on the same basis as founders, this may never be a problem.
We need to have lists of what everyone is working on, so that we know what's covered and what's being overlooked. Also, everyone should know where to go with new information or questions concerning any project.
There is now an MJK user name called TLIST which is to be used for sending the lists. By the Wednesday night before each meeting, please send TLIST a list of what you've accomplished and what you intend to have done by next month, on a fairly specific level. Please keep it compact, to avoid overflowing the number of lines that TLIST is allowed to have stored. On the Wednesday night before each meeting John Walker will pick up the messages and compile them into a document that everyone will get a copy of.
On our last bill from MJK,
35% of the computer time charge was for prime shift. We can save
substantial amounts of our working capital by not using the system
unnecessarily in the daytime. We are also running up charges for file
storage. If you use MJK at all, please pick up and delete your
messages.
The next (September) meeting is at 1:00 on Saturday, August 28, at
Jack Stuppin's house. There will be a demonstration of BitStik, a
frighteningly good graphics package on the Apple.
We've been making some attempt to keep to an agenda at these meetings without suffocating in Robert's Rules of Order. The responsibility for the agenda has fallen on the corporate secretary, Dan Drake. It would be nice to know what people now think about how the meetings should be run. Should we drop agendas entirely? Should we be more authoritarian in holding to the agenda? Should the agenda be changed in form?
In any case, if you have something that you think ought to come up at the meeting, please send a message to Dan Drake so that it can be brought up at a reasonable point on the agenda. If it seems too small a matter to put on the agenda, you can bring it up under new business.
Here is the normal skeleton agenda:
|
|