I was there. What killed Compuserve was an attempt to build an entire back-end infrastructure, software and hardware, de novo, using immature consumer-grade technology. It was a shocking thing to witness from the inside. The pivot from proprietary to open-standard was not the problem. The problem was trying to build a software ecosystem in which no one knew what they were doing and the OS (Windows NT) didn't work yet, and to populate a data center with almost-but-not-quite redundant commmodity compute and networking. Came time for the big relaunch ("WOW!") and for every person who succeeded in signing up for the new service, ten others tried and failed. No one could come back from that.
There is a guy on YouTube playing with SC-40s who has part of the TOPS-10 based stack, which I guess is what you mean. The Internetty stuff went by the umbrella name of HMI (Host-Micro Interface) and was guarded like the recipe for Coca-Cola. The boundary between the server and client was...fractal. Everything was about saving/spoofing bandwidth. The "browser", WinCIM, contained a huge library of UI element templates.
Systems Concepts may have had other prospects, but they had so much trouble getting the product out the door that Compuserve ultimately helped them finish it, licensed it (? exclusively), and produced all but a couple of the machines that went into production. I found a microcode bug in one of the first SC-built machines that caused the string-comparison instruction to return true whether or not; Gary Brown took that to SC and reported their reaction as "Damn, you caught us!", and that was kind of the end of that. You might have some stories to tell of the flight from X.25...?
I was there. What killed Compuserve was an attempt to build an entire back-end infrastructure, software and hardware, de novo, using immature consumer-grade technology. It was a shocking thing to witness from the inside. The pivot from proprietary to open-standard was not the problem. The problem was trying to build a software ecosystem in which no one knew what they were doing and the OS (Windows NT) didn't work yet, and to populate a data center with almost-but-not-quite redundant commmodity compute and networking. Came time for the big relaunch ("WOW!") and for every person who succeeded in signing up for the new service, ten others tried and failed. No one could come back from that.
I'm most curious what happened to the Compuserve server software, and could some of it be resurrected as an interactive exhibit?
Some paper was preserved by CHM: https://news.ycombinator.com/item?id=30848845
There is a guy on YouTube playing with SC-40s who has part of the TOPS-10 based stack, which I guess is what you mean. The Internetty stuff went by the umbrella name of HMI (Host-Micro Interface) and was guarded like the recipe for Coca-Cola. The boundary between the server and client was...fractal. Everything was about saving/spoofing bandwidth. The "browser", WinCIM, contained a huge library of UI element templates.
Found some photos of the SC-40 deal at bitsavers.
https://bitsavers.org/pdf/compuserve/
Ooh, and these SC-40s have compuserve tags and a stack of disks! I hope someone preserved the data.
https://onlineonly.christies.com/s/firsts-history-computing-...
AFAIK SC-40 were all made for CompuServe, and effectively by CompuServe
Systems Concepts may have had other prospects, but they had so much trouble getting the product out the door that Compuserve ultimately helped them finish it, licensed it (? exclusively), and produced all but a couple of the machines that went into production. I found a microcode bug in one of the first SC-built machines that caused the string-comparison instruction to return true whether or not; Gary Brown took that to SC and reported their reaction as "Damn, you caught us!", and that was kind of the end of that. You might have some stories to tell of the flight from X.25...?
Good times! I was 75060,264 and I used software called Golden Compass on os/2 to connect, post my messages, and retrieve new messages.