The man who made the Internet possible

Divemedic has been geeking out about home network stuff which has been a bit of a walk down memory lane - I once had a supervisor who turned down my request for a 2400 baud modem "because you can't read that fast".  You see, at 1200 baud, a text file comes down at around one line per second, and a fast reader can read that fast.  Sorry, Borepatch, no speed for you.

And quite frankly, nobody in the late 1980s got much speed.  We had big VAX computers and the standard wisdom was that "You'll never get more than a megabit per second out of TCP/IP".  And it was standard wisdom because nobody did.

Enter a fellow named Van Jacobson.  He looked at the source code for popular TCP stacks and decided that it could be improved.  WAY improved.  TCP wasn't implemented in the operating system, as an OS service; rather, it was implemented in a user application.  The problem with this, thought Jacobson, was that every time you wanted to send a packet the OS had to do a context switch to let the (user-land) data structure containing the packet go to the OS kernel.

Jacobson thought this was stupid, and that networking should be the job of the OS.  He hacked IP into the Unix kernel and all of a sudden not only were you getting more than a megabit from TCP, a single Sun Workstation could saturate a 10 Mbps Ethernet.

IBM basically said that this was because Ethernet was stupid and primitive, and IBM's Token Ring network enforced bandwidth sharing.

Enter Jacobson again.  He defined a mechanism called "Slow Start" which he added to his hacked TCP code. Basically, whenever you started to send data, you would start slowly and increase the speed gradually until you hit congestion.  What he found was that not only would his Sun workstation saturate a 10 Mbps Ethernet, but 5 or 6 Sun workstations would saturate that same network while gracefully sharing the network between them.

So why do I say that he made the Internet possible?  Imagine a bandwidth-constrained Internet - constrained either because Internet Protocol was too slow or because more powerful computers hogged all the bandwidth.  There would be no Tik-Tok, no Youtube, no Netflix, no Zoom.

And it instantly made Sun Microsystems into the Ferrari of computers.  As Networking took off, Digital Equipment Corporation couldn't compete with lower cost workstations that were much, much faster.  That was it for VAXen, and for IBM's Token Ring, and for ISO's Open Systems Interconnect protocols, and for Asynchronous Transfer Mode.  The problems that these technologies were designed to solve were suddenly obsolete.

All because of Van Jacobson.

Not many people remember this time of massive upheaval, but we all owe Jacobson a huge thank you.