How engineers at Digital Equipment Corp. saved Ethernet

Spanning Tree Protocol and Ethernet’s “Magic”

• Several commenters praise spanning tree for making Ethernet robust for non-experts, enabling ad‑hoc “extension-cord-style” cabling without physical disaster.
• Others note that STP seems magical at first but becomes straightforward once you see it as building an acyclic graph; one comment walks through a simple loop-detection example.
• Optimizing STP reconvergence is considered hard and “magical” even by technically literate participants.

Failures, Fire, and Power Strips

• Multiple anecdotes involve equipment “magic smoke”: blown capacitors from bad power and brownouts; lightning damage; misused passive PoE.
• A recurring theme is people misusing power strips: plugging a strip into itself, or yanking cords instead of using power buttons, sometimes destroying expensive AV gear.
• There’s debate over whether outlets in US/Europe typically have switches and how people actually power devices off.

Newer Layer‑2 Protocols (SPB, TRILL, EVPN)

• One commenter claims 802.1aq Shortest Path Bridging is a more robust, secure replacement for STP and criticizes how long it took to appear.
• A reply counters that SPB is effectively a “dead letter” standard; operators that would need it tend to use BGP EVPN or other L3 solutions instead.
• TRILL is mentioned as an earlier IETF alternative; IEEE’s later SPB decision is framed as a course correction, but both are said to have limited real-world uptake.

Incremental vs. Revolutionary Innovation

• A parallel is drawn between Ethernet bridges and a philosophy of small, local, incremental improvements leading to big wins.
• Another commenter pushes back, arguing network-layer bridges were a qualitatively new device, not a small tweak.

Ethernet Speed Evolution (10 → 100 → 1000 Mbps and Beyond)

• Several comments explain the long life of 10 Mbps: it was “good enough” relative to small disks, modems, and printers, and faster gear was costly.
• The 10→100 jump required new cabling (Cat 3→Cat 5, plus coax phase-out), making upgrades expensive and slow.
• The 100→1000 jump was faster because it could reuse existing Cat 5 and cheaper PHYs.
• Some note that consumer gear still mostly tops out at 1 Gbps despite affordable 10G-capable components; others argue 1 Gbps is still overkill for most.

Switches, Segments, and Operational Pain

• Early switches did not immediately mean one port per host; they often fronted shared coax segments.
• Over time, as costs fell, networks moved from shared media → hubs → per-host switched ports for reliability and performance.
• Older coax-based Ethernet is remembered as fragile: a bad terminator or user tampering could take down entire floors, making modern switched topologies a welcome tradeoff.

Historical and Practical Nostalgia

• Commenters reminisce about DEC hardware (e.g., Tulip NICs), early SynOptics/3Com gear, and the rapid shift from coax to twisted pair.
• There are memories of juggling DOS network drivers and memory extenders to run multiple protocols, and of early hubs saturating under modest traffic.