PLNOG 21: Ron Broersma - Historical_Perspectives_on_Computing, Networking, Security
0 likes55 views
This document summarizes the state of computing and networking in 1976 from the perspective of Ron Broersma, a scientist at the U.S. Navy. In 1976, mainframe computers dominated, TCP/IP was still being developed, and there was no internet, web, or smartphones. The ARPANET used the NCP protocol and 8-bit addresses to connect a small number of sites. Unix version 6 was an early operating system that came with source code. By the early 1980s, TCP/IP had replaced NCP as the standard, and the transition to networking of networks enabled the internet to emerge. Security incidents in the 1980s demonstrated the need for protection where the early networks had been open research environments.
PLNOG 21: Ron Broersma - Historical_Perspectives_on_Computing, Networking, Security
- 1. Historical Perspectives on Computing, Networking, Security Ron Broersma, Scientist, U.S. Navy Chief Engineer, Defense Research and Engineering Network
- 2. ARPANET July 1976 I was here at the Naval Undersea Center (NUC)
- 4. Underwater communications worldwide Jim Creek VLF Naval Radio Station, 1953 24.8 kHz at 1.2 Million watts
- 5. The state of things in 1976 • It was the “mainframe era” • Before the IBM-PC, or any PCs – some hobbyists had IMSAI or Altair micro-computers. • TCP/IP still being developed and tested • Unix version 6 (before BSD and any other variants) • No domain names or DNS • No email • No ethernet • No “open source” – Actually, no source available at all (proprietary) • No Internet, no web, no social media • No cell phones
- 6. Computing culture in 1976 • Frustration with the “mainframe priesthood” – users wanted more control and access • 1975 – “Computer Lib” and “Dream Machines” published – envisioned things like “Hypertext”
- 7. The state of networking in 1976 ARPANET • Protocol: NCP (Network Control Program) – 8 bit addresses • 2 bits for host part, 6 bits for IMP number • in that era, computers were expensive mainframes and nobody could envision a site with more than 4 computers. • My IMP was node #3 – addresses separation used a “/” – our address was 0/3, which could be shortened to just “3”. – Later, our IP address became 10.0.0.3. (Class A - no CIDR) • One used the “BBN 1822 protocol” to connect a host to an IMP. – interfaces types were “LH”, “DH”, or “VDH” • Circuits between IMPs were 50 Kb/s wideband analog modems – very large: you could fit 2 of them in a rack
- 8. The front panel from IMP #3 BBN Model C/30 San Diego, 2012 Boot from cassette tape
- 9. Connecting your terminal • Terminal IMP (TIP) – serial connection to IMP for terminals (teletype) • 110 or 300 baud modem for remote access (later 1200) • Later called Terminal Access Controller (TAC) • Centralized authentication used a newly invented protocol “TACACS” • You were sent a “TAC card” in the mail with your userid and password. ASR33 TTY
- 10. Our first computer on ARPANet • DEC PDP-11/40 – 16 bit architecture – Dual 8” floppy disk drives – 2.45 Mbyte disk drive – Added 60 MB SI disk drive • Ran the “ELF” OS • NCP protocol to ARPANet … and then we heard about “Unix”
- 11. Getting Unix • Version 6 Unix - $20,000 • Comes on 9 track tape or disk pack • Comes with instructions from Ken and Dennis
- 12. Unix Version 6 Instructions Page 1 Page 9
- 13. Running Unix on PDP-11 • Booting – Used lights and switches to enter boot code – That booted from 8” floppy disk – Which then booted from the 60 Mbyte Systems Industries disk drive • Recovery from crashes – icheck, dcheck (no fsck) • Configure and build kernel (/usr/sys/run, mkconf) – compile in device drivers, cpu options, interrupt vectors • Challenge (disaster recovery scenario): – Run Unix on just the RX01 (8 inch) floppy drives • 256 Kbytes each • One for root, one for swap • Had to rewrite the rx driver – Included TCP/IP and telnet server
- 14. Unix was evolutionary • Came with source code – You could read the kernel source and understand it • You learned good C programming techniques from this – You could enhance the system – You could augment it (add networking) • Hierarchical filesystem that was simple to understand and use • setuid bit • Shell, and simple commands that could be interconnected with pipes
- 15. Unix in the 70’s • Version 6 (1975) • PWB/Unix (1977) – rje, sccs, make, and other tools for programmers • Seventh Edition (1979) – Bourne shell, awk, tar, uucp, and much more • Unix/32V (1979) – Runs on a VAX (32 bit arch, virtual memory) … and then we heard about “BSD”
- 16. We tried BSD • We tried 2BSD on PDP-11s, and 3BSD on our new VAX-11/780 • 4BSD (late 1980) – delivermail, curses library and termcap – slow – VMS was much faster • 4.1 BSD (‘81) – Numerous fixes, much faster – cool new things, like #! – Later versions (leading up to 4.2 BSD) included TCP/IP networking (4.1a, 4.1b, 4.1c) • 4.2 BSD (’83) – slow • 4.3 BSD (‘86) – great performance
- 17. 1982 – me running 4.1 BSD on a VAX-11/780
- 18. Transition from NCP to TCP: There was a plan
- 19. Transition to TCP/IP • There was a published plan • Both protocols supported during transition • There was a deadline: Jan 1, 1983 – NCP was turned off • Organizations were highly motivated to transition – Does this lesson apply to IPv6 transition? • New era: “packet switching” – no longer needed to connect host directly to IMP – network of networks ! Internet – Router: enable forwarding on Unix, run “routed” or “gated” using EGP (precursor to BGP) • I am ASN #22 – Ethernet: yellow cable and vampire taps • Most exciting test: “telnet ucl”, get login prompt
- 20. Tracking names and addresses • All computer names and IP addresses were kept in the “hosts file” – HOSTS.TXT – one would have to download it from the NIC using FTP every so often • save it in /etc/hosts – there were also some unauthorized versions that contained additional names and aliases (many were humorous) • Soon, all the good names were taken and conflicts started, so there was a need to have multiple name spaces (domains) – DNS was born (specified 1983, BIND in 1984) • my email address changed from ron@nosc-cc to ron@nosc.mil
- 21. Our first security incident August 1983 • Password guessing from UCLA over the ARPANET • No longer a friendly research playground where you trusted everyone. • Everyone needed to start locking their doors. TRS-80 NOSC
- 22. Early security references • The “Orange book” (1985) – “Trusted Computer Systems” – part of the “Rainbow series”.
- 23. more on security • Additional notable incidents – Morris Worm (Nov 2, 1988) – El Griton (1995) • Fighting back: – “hack back”
- 24. autodialers and UUCP • Before the days of Hayes modems, where you could use “ATDT” commands to dial a phone number • Build a driver to pulse a phone line to do dialing • Use UUCP to send mail and news over dialup – huge phone bills • If your dialer timing wasn’t right, you could get wrong numbers – oops
- 25. Sharing with Friends • Connected neighboring Universities via Serial Line Internet Protocol (SLIP). This was before PPP existed. Examples: – UCSD via 2.4Kb dialup modem – Univ of Hawaii (9.6Kb across Pacific Ocean to Hawaii, then SLIP over dialup to Honolulu) • We provided authoritative DNS for numerous large domains. • This all changed after NSFNet arrived (1988) and connections were more generally available. • 1992 – Internet becomes mainstream
- 26. Summary • Birth of the Internet led to many subsequent innovations – Multiple transitions along the way • From NCP to TCP • From circuit switching to packet switching • From open and friendly research environment to something needing protections • From government to commercial • Unix and derivatives revolutionized computing – From closed proprietary systems to open source and portability • History can help explain why things are they way they are, and can provide valuable lessons.