Into the Telephony Subsystem

3 Replies

It is spring of 1998 and I was starting off at Edify, the company that would be, from one perspective, my place of employment for the next twelve years.

Stability, in Silicon Valley?

During that dozen years the company would be bought, spun off, bought again, merged, then bought one more time before laying almost the whole staff off… after which the company that did that would be bought again, capping off my resume where I literally ten employers listed over 22 years, most of which do not exist at all or in any recognizable form today.

Edify Corporation

Despite being a successful company whose software was used by a large number of big name companies, there is little or no record left online about it.  At least Global Village managed to be worth a page on Wikipedia.  I can dig and find things about Big Island and the YoYo.  Edify though, that is just some smoke in the wind now.  If you didn’t know it was there you would never know it was missing.

That is Silicon Valley.  A lot of medium size successful businesses come and go without a ripple, and the fact that a few hundred other people have jumped on the name Edify since then doesn’t help.

That roller coaster of a ride started with just Edify, which had just gone public about a year before I signed on.  As I mentioned last time, Edify was in the enterprise software domain, which meant better pay and some stock options just for coming on board.

The company had been founded as Reach Systems around 1991 by a group of engineers who had come from the Rolm PhoneMail group, and a hearty set of old school devs they were.  They were serious engineers in a way that a lot of Silicon Valley no longer is, rooted in the foundations they had learned dealing with embedded systems where you had to fill out a justification form to use up any of the precious RAM those boxes had back in the day.

Having seen the potential for computer telephony and what would later be called IVR systems, they sought to carve out a new market niche they called “self service” that would allow average department level administrative people build and update phone apps.  By 1994 they were headed to market with their new product.

Email was very important in 1994

This is a recurring theme for me, the idea that a company can make an easy to use development environment for telephone apps… I totally forgot to mention that Cypress Reseach had such a vision at one point, and I had tinkered with Prometheus’ speech app development environment way back in the day… that will lead to sales because people really want to develop their own custom apps.

Spoiler: No they do not.

And so, as happens in every iteration, the company created a professional services group that would use the software, called Electronic Workforce, or EWF around the company, to build solutions for customers.  Dumb companies sell technology, smart companies sell solutions, or something like that.

Anyway, that was a viable business plan and by the time I joined they had some big customers like MCI, Kodak, and Sears, along with a host of smaller companies and an close ecosystem of value added resellers, who also offered professional services and app hosting and all of that sort of thing.

Edify, which was the third name of the company… and I have forgotten the interim name that was in there between Edify and Reach… set off to build such beast on IBM’s OS/2.

They can be forgiven that choice because, at the time, for a serious enterprise app you could use that or some flavor of UNIX.  They couldn’t know how badly IBM would botch the OS/2 marketing plan.  When you proudly declare that your operating system runs Windows software better than Windows, you have effectively told everybody they don’t need to support your OS because they can just do Windows apps and you’re covered.

But that is another story.

EWF was broken up into subsystems, each of which handled things like connections to databases to look up data, connections to terminal emulation to screen scrape data, connections to home brewed DLLs to do any task you wanted, along with a bunch more that were focused on data retrieval.  These were all connected via an object bus written early in the development of the product, and which was so robust that it is still the core of whatever is left of the product today.

Then there were the subsystems that connected to the outside world, and the primary amongst those was the telephone subsystem, which handled inbound and outbound dialing.  Edify had gone with Natural Microsystems for telephony hardware support, so we had a host of ISA, then PCI cards to deal with for configuration and support.  You can still  find many of the boards I worked with… the AG-4 and 8, the AG-T1/E1, the AG4000… for sale on eBay or Amazon.

An NMS Alliance Generation (AG) T1/E1 ISA Board

Voice over IP eventually replaced all of that, though not without a period of extreme awkwardness that we’ll get to later, but at the time you needed a card in a slot that could hook up to a telephone line.

Natural Microsystems (NMS), which was based near MIT… there was a nexus of speech and telephone companies around both MIT and Stanford/SRI that were off-shoots of the research being done at those schools… was the more open alternative to Dialogic.  Dialogic had more market share, but NMS was more open, easier to work with, and was less focused on UNIX, so Edify had opted for the latter.

And into this mix in the spring of 1998 I wandered, starting the same day as another member of the QA team who would go off to work on the terminal emulation subsystem.  At one point, who was the more senior of the two of us would become a minor issue when we moved buildings and the VP of engineering let us pick out cubes based on seniority.  The conflict was resolved because there was only one window cube left by the time we were up and I didn’t want to sit by the window.  He did, ignoring the fact that it was also the cube next to the VP’s office, and when you sit next to the VP you effectively become his admin, so people were always asking him if the VP was in or where he was.  I am pretty sure I won on that deal. (Always sit as far from the boss as practical.)

Edify was in the midst of transitioning off of OS/2 and onto Windows NT.  They had shipped the 5.0 version of the product on Windows, but that first release only had a subset of the OS/2 features, so there was work to be done.  The 5.1 release was just being launched as I signed on, and I was there to work on the 5.2 release, which was slated to be feature parity with OS/2.

I was hired because I knew things about ISDN from Big Island… though absolutely nothing useful or applicable, but I knew SOMETHING, which was apparently more than the current team… and because I had experience with Windows NT.  As mentioned, I had to install OEM configs of various versions of Windows back at Cypress Research.  I could install Windows NT 4.0 in Japanese… also not particularly useful… but I knew how to work with the device manager, which was.

And my new boss was a co-worker from Global Village.  I also had another former GV co-worker there as well.  The easiest way to get a job in tech is to know somebody at the company who will recommend you.  So I was hired and joined the telephony team.  Engineering was broken up into groups based on which subsystem you worked on, and I was in the telephony club.

Once I had been brought on board, had been given the telephony history talk… which I would later give myself to new members of the team… went to Edify training… we had a training team to teach people how to develop apps in EWF… and could install, configure, and get a simple “Hello world” level telephony app running reliably, I was given my first assignment.

I had to get the NMS DTI-48 configured to run with EWF.  The DTI-48 was an older NMS board, one that had long been superseded by the AG range of cards, but some of our OS/2 customers had them and we were worried they wouldn’t upgrade to Windows NT until we supported that card.

Here’s the truth.  I could never get that card to work.  It was a two card solution with all the signal processing on one board and the phone interface… two T1 lines, the name standing for something like “dual trunk interface, 48 lines” if I recall right… that was old enough that the NMS tech I ended up working with on it hadn’t ever seen one either.

But no customer ever made their order contingent on us supporting that card either.  So I spent a couple of months learning about NMS configurations and banging my head against this one particular setup.  Nothing like completely failing on your first assignment, though at least I had other tasks to work on while the machine in the corner of my cube with the DTI-48 just sat there, sullen and brooding, a menace I tried to ignore.

I had been given a double size cube… they had originally been planned as double occupancy cubes for pairing… which was not yet a common concept… but people hated that so the telephony people got them because it let us set up a mini-lab in our cube with a couple of PCs… 200 MHz Pentium Pros all around at that point… with the requisite 20″ CRT monitors along with an 8 channel analog phone line simulator and a few phones plugged in for good measure.  You needed at least three if you were going to test transfers.

So I puttered along, not exactly setting the world on fire, but neither being grossly incompetent.

Then my boss came along and asked me if I wanted to work on something new.  In Silicon Valley, being offered a chance to work on something new is often a lifeline, a chance to leave whatever you’ve messed up behind you, to hide the evidence, bury the skeletons, and start a fresh new life withe a fresh new project.  It is like witness protection, but you don’t have to move to a new town.

So I said yes.  That DTI-48 had bested me… though in hindsight, the whole task might have been an initiation ceremony, as nobody else ever got the damn thing to run and it was eventually forgotten about… and I was ready to dive into something different.

I and one of the sharp new developers drove up from out Santa Clara offices to Menlo Park, to an office on Willow Road, not to far from what was once the headquarters of Sun Microsystems and is now where the main Facebook campus is located, in what was at the time kind of a dicey part of town.  Rising real estate prices had yet to gentrify that part of Menlo Park and East Palo Alto.  Hell, there were still “bad” parts of Mountain View back then.

We were headed to the offices of Nuance Communications, one of those SRI spin-offs that you can find in the valley, where I was going to learn all about speech recognition, a technology that would dominate my career for the next few years.

The story so far:

3 thoughts on “Into the Telephony Subsystem”

  2. heartlessgamer

    Thanks for making me feel old(er) by having read this. 15 years in telecom now (all for the same company) and the only thing I know outside of my time in the Air Force.

    Like

    Reply ↓

Voice your opinion... but be nice about it...