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 am from a place where CompuServe was just something you read about the way you read travelogues or otherwise reports from far away country, unfortunately.
I encountered data about System Concepts due to being interested in PDP-10 history and so on.
It was one of multiple successors to PDP-10 that tried to capitalize on existing user base after Digital decided to utterly cancel all 36bit work in favour of VAX (legends claim the last prototype ended up in millpond).
Some started even earlier, trying to compete on speed or other attributes (like Foonly)
System Concepts is the ones who ended up with CompuServe as client/effective owner from my memory. XKL was somewhat successful to the point they are the reason why last known PDP-10 was made after closure of last VAX manufacturing line, but nearly all of their systems was for internal use in their network gear
There were options for migrating from PDP-11, not PDP-10.
PDP-11 was a 16bit byte-addressable minicomputer that was extended to 22 bit physical addresses through MMU. VAX took PDP-11 and extended it into 32bit flat addressable paged system.
PDP-10 was 36-bit, word-addressed with 18bit addressing (plus segment extensions in KL-B that were only really implemented in 3rd party follow-ons, a mainframe-in-all-but-name.
VAX would mean starting from zero unless nearly everything was written very high-level, which is doubtful given the optimizations needed to run so many simultaneous users.
It had more to do with the ecosystems and the corporate relationship. The -10s came from a earlier perspective. It was open source before "open source". You bought the software and you got the source, because of course you did, you couldn't even install it otherwise. Compuserve tracked TOPS-10 up to version 6.01, with their own additions and patches: far-reaching but well encapsulated. They could not track 6.03, so they "forked" it, to use the modern word. The question of support was always fraught. Compuserve offered all of their changes to DEC, who mostly sniffed at them. But from DEC's perspective, one of the major purposes of the VAX platform was to lock all that kind of thing down and go to a support model more like that of IBM.
Another part was pretty vicious infighting between PDP-10 (already sorta on thin ice from start due to fuckup that was PDP-6) and PDP-11/VAX teams, and VAX team won it pretty decisively - if in part through consolidation efforts.
EDIT: Another major difference was how PDP-10 was from an era where everyone with a (under)graduate students to throw at the problem was making custom modifications to hardware.
Like some early multiprocessor PDP-10 being done with hand-soldered MP bus done by students.
30 years later, 76344,2701 is still burned into my brain. I used a DOS text-based program whose name I don't remember anymore. I know I used OZCIS.exe at some point, and Aladdin with GEnie.
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...?
I am from a place where CompuServe was just something you read about the way you read travelogues or otherwise reports from far away country, unfortunately.
I encountered data about System Concepts due to being interested in PDP-10 history and so on.
Profound apologies -- on the basis of one of your earlier comments, I took you for someone with the same initials.
Was the whole SC endeavor to avoid buying PDP-10s?
It was one of multiple successors to PDP-10 that tried to capitalize on existing user base after Digital decided to utterly cancel all 36bit work in favour of VAX (legends claim the last prototype ended up in millpond).
Some started even earlier, trying to compete on speed or other attributes (like Foonly)
System Concepts is the ones who ended up with CompuServe as client/effective owner from my memory. XKL was somewhat successful to the point they are the reason why last known PDP-10 was made after closure of last VAX manufacturing line, but nearly all of their systems was for internal use in their network gear
DEC quit making them in 1983.
Then VAX would have been an option, especially since there were migration tools from PDP. And VAXen cluster for scaling.
There were options for migrating from PDP-11, not PDP-10.
PDP-11 was a 16bit byte-addressable minicomputer that was extended to 22 bit physical addresses through MMU. VAX took PDP-11 and extended it into 32bit flat addressable paged system.
PDP-10 was 36-bit, word-addressed with 18bit addressing (plus segment extensions in KL-B that were only really implemented in 3rd party follow-ons, a mainframe-in-all-but-name.
VAX would mean starting from zero unless nearly everything was written very high-level, which is doubtful given the optimizations needed to run so many simultaneous users.
It had more to do with the ecosystems and the corporate relationship. The -10s came from a earlier perspective. It was open source before "open source". You bought the software and you got the source, because of course you did, you couldn't even install it otherwise. Compuserve tracked TOPS-10 up to version 6.01, with their own additions and patches: far-reaching but well encapsulated. They could not track 6.03, so they "forked" it, to use the modern word. The question of support was always fraught. Compuserve offered all of their changes to DEC, who mostly sniffed at them. But from DEC's perspective, one of the major purposes of the VAX platform was to lock all that kind of thing down and go to a support model more like that of IBM.
Another part was pretty vicious infighting between PDP-10 (already sorta on thin ice from start due to fuckup that was PDP-6) and PDP-11/VAX teams, and VAX team won it pretty decisively - if in part through consolidation efforts.
EDIT: Another major difference was how PDP-10 was from an era where everyone with a (under)graduate students to throw at the problem was making custom modifications to hardware.
Like some early multiprocessor PDP-10 being done with hand-soldered MP bus done by students.
It was a wild place
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.
30 years later, 76344,2701 is still burned into my brain. I used a DOS text-based program whose name I don't remember anymore. I know I used OZCIS.exe at some point, and Aladdin with GEnie.