Dubman.com/narratives

Microsoft Works 4.0 and the birth of MSN

Jonathan Dubman circa 1992

Teammates jump ship to MSN

After we shipped MacWorks 3.0 in 1992, a contingent jumped ship to join Marvel, the new “Microsoft Online”, which eventually became MSN. What became Windows ‘95 was supposed to be Windows ‘93. We called it “Chicago”, the city of my youth. Online was obviously the future. I’d been online for 15 years at this point. I would have had the coding skills to a build an early web browser from the ground up if assigned the task, but that wasn’t my job. I was working in Apps.

When teammates left to join MSN, I stayed on to write Mac apps in the Consumer Division, out of loyalty to my team and my product. I thought it was the right thing to do for the company. I took one for the team, and remained on that team for another two years to ship Works 4.0 for Macintosh in 1994, just as the early web was being born.

Joining the online cohort would surely have accelerated my career. Windows was really taking off, the Mac was in a slump, and it was not the most strategic move for my career to remain on a Mac-only product at Microsoft.

A smaller vision prevails

Charlie English, the Works PM who jumped first to MSN, and then off Tiger Mountain (with a paraglider!), had a much broader, far-reaching, web-like vision than what MSN actually became, which was a misguided attempt to imitate AOL. We got updates on this at a different bar every week. For posterity, I kept an internal email from May 14, 1993 entitled “Formation of On-line Services Group”, reflecting the vision that Bill Gates and Russ Siegelman had settled upon against Charlie’s wishes. This may otherwise have been lost to history as it is not showing up in a web search.

BillG writes: “The immediate mission of the group is to launch a branded Microsoft consumer service, concurrent with the Windows Chicago release. This will be a dial up service and the client will be included in the Chicago box. Current thinking is that the content will include online registration, PSS, MS mktg and developer info, national news, sports, weather, financial, games, email, chat and bbs.”

In contrast, Charlie thought (as I agreed) that a far better approach was first, to ship a more generic client-side browser (like the web browser that eventually became IE 1.0 in the Plus pack), and then, to jump in the business of selling server software and tools to give everyone and their businesses an online presence, which everyone will want to have. What a crazy idea. When his vision was met with a big shrug by executive leadership, Charlie quit Microsoft, taking his brain with him, to open a wine bar and eventually found Dinerware.

A few of us got together on the side to form our own micro cloud computing initiative for our own needs with rented rack space in a closet at the Westin in downtown Seattle. It once hosted Dubman.com.

A new text editing engine for Works

Meanwhile I was still fully committed to the Mac app Microsoft hired me for, which, it’s worth noting, turned a healthy profit as a business, unlike hip new MSN, which burned through billions chasing barely-monetizable eyeballs.

For Mac Works 4.0, the rational decision was made to switch to the new “Quill” text editing and layout subsystem based on the Word core, in a partnership with Chris Mason, whom I knew as the development manager for all of Word during my internship.

Upgrading the word processor engine did make sense architecturally, but in hindsight I’m not sure if it penciled out, because it was a huge deal of work just to get to feature parity let alone add new features, and customers didn’t care at all what was happening under the hood.

Cinderella? What’s up with the vision, doc?

Upgrading the word processor engine was actually less of an impact than another fateful decision for Works 4.0 for the Mac. A slew of additional requirements were imposed on us top-down by the organization to support a new Microsoft technology called OLE, for Object Linking and Embedding, the kind of name only its inventors could love. I went through the docs, and was not impressed. As conceived, OLE didn’t seem to target real user needs, and would yield sluggishness and bloat even before adding any ambitious new features.

To make matters worse, OLE would force a file format change on existing users, while manufacturing the need for us to engineer and/or commission another round of file converters. This would take time for us and for our users, and cost us money. All for what?

After delays in completing version 3.0, the PM’s picked “Cinderella” as our 4.0 code name. This time, the magic would end at midnight, our carriage turning back into a pumpkin! Surely, that would motivate us! It seemed almost cynical to me, certainly not inspiring.

Our spec suffered from a scattershot approach to adding features with little to distinguish us in an increasingly crowded field beyond the fact that we were there first. Being first would have limited lasting value without some kind of network effect to build on that user base, or a reason for existing users to upgrade.

A slate of patronizing “wizards” (a new concept basically pioneered by our business unit) would offer to help novice users do tasks that we were simultaneously claiming were inherently easier than ever to do in our app. This was an incoherent marketing message and didn’t feel right to me. Offering that much help would make it all look harder than it really is, while taking precious engineering time to implement. I also didn’t think treating our experienced users (the many potential upgraders) like novices would be received well by our installed base. Notably, it was our business unit that also produced the infamous Microsoft Bob.

To me, it seemed like it would make marketing (the job of Melinda’s group) easier if the new version of Mac Works were clearly better than the competition at something both new and useful. And despite all the big talk in the “Vision Document” I wasn’t sure what that something was supposed to be for version 4.0. It was a pile of unrelated new features with few standouts.

In an alternate universe, we could have pioneered the LAN-based real-time collaborative editing idea I describe below, à la Google Docs, which would have been instantly useful for schools, families and small office teams alike, and that feature would made a splash in the early 90’s.

But back in the real world, we would focus our efforts on enabling features like dynamic chart updates (which were still slow) for the type of people who printed all their documents. We built these features that none of our customers were asking for to support Microsoft’s grand strategic vision at the time.

Anyway, my next job would be to overhaul a word processor I had just spent two years overhauling before in order to accomplish all this, a Sisyphean task.

A new, integrated drawing module

I also rewrote a vector-based drawing module, sort of like a mini Microsoft Visio or Adobe Illustrator, which was deeply integrated with the word processor. I got live-editable multi-columnar text to flow around irregular graphics, on ancient platforms with slow processors and tiny memory. It did a lot of stuff Word couldn’t even do (actually not the first time I had done that!)

Everything worked well, but it was surely not the optimal use of our time. While we were working on all this, the Apple subsidiary called Claris kept innovating, and started visibly beating us on simplicity, discoverability, usability and performance - the same stuff I would have preferred to focus on for our latest version. This was eating into our market, which was both validating, dispiriting and motivating all at the same time.

Wanting to invent Google Docs in 1994

Collaborative real-time editing over a network (like Google Docs, but a full decade before Google Docs) kept repeatedly popping into my head as feasible and hugely compelling, but it was out of scope for our project. I had experienced real-time text messaging over BSD sockets at Berkeley in the mid 80s, and I had even implemented two-way real-time online text messaging on my own in 1984 (two decades before Google Docs) as a side project I never even bothered to brag about on my resume. It would have worked just fine on PCs or Macs in the early 90’s.

I didn’t understand at the time why it wasn’t obvious to everyone that collaborative editing and real-time messaging for teams would both be super useful. Office workers on one of the new LANs Microsoft was so excited about wouldn’t even need an internet connection to collaborate locally. And there was no inherent reason collaborative editing and real-time messaging couldn’t be performant over dialup. Editing is just a stream of discrete human-initiated actions, and humans are slow typists who basically do one thing at a time. I should have just built it, a decade before Google Docs, but Microsoft had no “20% time” for side projects, and I didn’t either. More like minus 20% time. SkunkWorks was not to be.

Microsoft was an industry-leading juggernaut, but from deep inside, I saw many aspects of Microsoft’s corporate culture that worked against innovation. Too much stress and territorialism, too many silos, no time or encouragement for side projects, short-term thinking despite all the grand visions, fear of disrupting existing revenue streams, and a posh but uninspired car-dependent suburban office park that prioritized comfort over collaboration. I did appreciate the forested hiking trails, a fine place for a moving meditation on all the stuff we were doing wrong.

Actually inventing: a patent for spell checking

Along the way, I worked with the Proofing Tools Group on Word and scored a patent for optimizing spell checking. I invented the approach that tracked which parts of the document had been edited or not (the “clean bit”), and avoided re-checking the unchanged and unedited portions. Things got interesting with mutable custom dictionaries. I worked alongside the Word team to deliver this functionality on multiple platforms.

I think the work I did on this may have outlived everything else from this period. I see continuity from this adaptive spell check work, through change tracking and grammar checking, to the AI-powered rewriting features Word has today.

Microsoft, and my career, at the crossroads

Windows 3 was a big success in the early 90s, but it was Windows ‘95 that made Microsoft a cultural phenomenon of its own, coincident with the emergence of the web and, separately, widespread voice-only cellular phone service. Microsoft was hitting its stride in its initial ascendancy during this period, and it was a heady environment. The broader tech industry was on the cusp of becoming central to modern life.

As we completed Mac Works 4.0 in 1994, I was quite depressed about losing my father the year before. Microsoft had forced me to cancel my last summer vacation with him on the coast of Maine to stay and write code in Redmond. I remember driving home in the rain back then, skipping my exit, and heading for a lone patch of blue sky over West Seattle. As I stood on Alki Beach with my Motorola Micro-TAC flip phone, my father and I mused about a future world of 2010, a future we figured would include mobile broadband in your pocket, a world he rightly said I would live to see. He told me about his friend working on the AT&T “Hobbit” processor, a low-power RISC chip that was used in prototypes for the Apple Newton. A future filled with what would one day be called smartphones seemed inevitable to us, even if was not imminent. We talked about how 5 years (1998) was still too soon for all the pieces to be in place, but 15 years (2008) was surely enough.

He saw me achieve some success, buy a house in Seattle and get on a good path in life. I had been able to buy him Apple’s top of the line PowerBook 170, the very first, as a 60th birthday present, which made him ecstatic. It felt like a time of transition, time to move on from Works and productivity software in general and find something more exciting and new. Apple sadly seemed to be fading.

What next: Online? Multimedia?

I looked into some early, early Microsoft mobile teams and office real-time communications teams that were years ahead of their time in some ways, but each was on a track I felt was so off the mark that it was unlikely to make it, and it didn’t feel right to join a project I thought was vaguely doomed just for the experience. There was a “Microsoft At Work” team starting its ill-fated effort to coordinate office equipment like FAX machines using Windows. I thought they were trying to solve the wrong problems with new technology. Coordinating FAX machines? People down the hall were about to make them obsolete. I wondered why they weren’t attempting something more innovative with mass-market appeal, like a contacts app with visual voicemail.

A rant about OLE

The Works diaspora was well represented in all the online projects. I interviewed with the “Blackbird” team that started out trying to implement what was essentially Apple News - in 1994. This seemed compelling, but they wouldn’t hire me because I didn’t know (or maybe didn’t care) enough about OLE, a colossal grab-bag of needless abstractions I didn’t even think they should be using in the first place, for which the market demand was zero.

The demand for OLE came from the top down, from BillG and his lieutentants who worked on it, who so thoroughly bought into this essentially defensive strategy to protect Windows’ newfound market domination that they were blind to the fact that OLE was like an anchor on any project, a huge dead weight preventing forward progress. What real world scenarios required this approach, and what was the opportunity cost for Microsoft to have so many smart people creating so much complexity and additional work for themselves and everyone else at that critical time?

I had close friends who worked on or with OLE. I had worked with it myself when we were required to integrate OLE into Works for the Macintosh, which was truly nuts because the Mac demand for Windows-based OLE was less than zero. With the slow computers of that time, app startup time and runtime performance were especially critical, and the competition was already faster before we starting adding all this stuff nobody wanted. My own development experience with OLE was like pair programming with a black hole.

I thought then (and now) that OLE itself was a technical solution in search of a problem, that distracted from the actual problems. The “Blackbird” visions of custom authoring tools and rich interactive OLE controls seemed like a huge waste; I thought the news belonged in a general purpose browser, not some complex custom client; all users needed was to customize their news feed and view the news, with embedded and captioned photos. I thought the true value of news is that it’s news, that complex print-like layouts were familiar but overrated, and all you really need is basic markup and plumbing for all the dynamic content flowing in from a limited set of sources. Why all the OLE stuff, was there some latent demand to embed live Excel documents in a news article in V1 of this platform? And if you want to monetize media in the age of slow dialup internet, make sure you have something multi-platform that’s as simple and efficient as possible, not something bloated and slow even when running on dev boxes over Ethernet.

It eventually dawned on me that at the end of the day, the problems OLE was designed to fix were not the customer problems, they were those of Microsoft itself. At the top level, BillG saw OLE as a way Microsoft could maintain and extend its domination as the Internet tidal wave was breaking on shore. That’s probably enough of a rant.

But I love maps!

I also interviewed with an Atlas/map team, a space I always thought had enormous potential. It was right up my alley. I’d been making maps and mazes since drawing them in chalk on the floor at age 3. I’d literally had a lucid dream of a scrollable, zoomable, vector-based Google Maps-like interface in the mid 80’s that I dubbed Metropolis, after the silent film, but I lacked the data to populate it. My walls were covered in maps at home and at work.

This would be in C++ with some 3-D graphics, which I’d already played with on my own while working on the math degree. I know would have spurred innovation on that Atlas team, especially since I was also dreaming of mapping destinations and businesses. But after stringing me along for months, in the final round, they chose another candidate, a match that didn’t end up working out for them… that’s how things go sometimes.

My next job

However, I soon scored another job in the new Interactive Media Division, which was kicking off all sorts of creative projects and collaborations to demonstrate the potential of Windows to a big new consumer PC market. This was an engineering role on Microsoft Bookshelf, a collection of multimedia titles (including maps!) produced in the same business unit as Encarta. Bookshelf was dual-platform, Windows and Mac, and it would be the next chapter of my Microsoft career.