Ποιά η σχέση της Θεωρίας στο σχεδιασμό του Διαδικτύου Γ.Κοροβέσης, ΙΠΤ ΔΗΜΟΚΡΙΤΟΣ 22 Ιουνίου 2010 ycor99ATgmail.com
Αναδρομικό σχέδιο: End-to-End Design J.Saltzer 1978, Naming and Bindings of Objects J.Shoch 1978, Inter-Network Naming,Adressing and Routing J.Saltzer 1982, On Naming and Binding of Network Destinations Saltzer,Reed,Clark 1984, End-to-End Arguments in system design Clark, Tennenhouse 1990, Architectural considerations for a new generation of protocols
Αναδρομικό σχέδιο: End-to-End Design (2) Clark 1991, Towards the Future Internet Architecture RFC 1287 Blumenthal, Clark 2001, Rethinking the design of the Internet Clark et al 2005, Tussle in cyberspace: defining tommorow's Internet T.Roscoe 2005 HOTNETS, The End of Internet Architecture
Αναδρομικό σχέδιο (3): OSI Programme Charles Bachman 1978 The OSI model of Layers H. Zimmermann (1980) The official OSI model Cerf, McKenzie, Scantlbury, Zimmermann1978, Proposal for an End-to- End transport protocol
Future Internet design McKeown, Girod 2006, Clean-Slate Design for the Internet John Day 2008, Patterns in Network Architecture: a return to fundamentals Louis Pouzin of CYCLADES
Η αξία της Θεωρίας Ιδέες στο μπλογκ t.blogspot.com/2007/07/blog-post.htmlhttp://ariadne- t.blogspot.com/2007/07/blog-post.html David Turner Church's Thesis and Functional Programming...”convergence of concepts... so to avoid having Intel Computing vs Motorola computing or Java vs C computing....” An Explanation of Computation Theory for Lawyers by PoIR in
TCP/IP problems area Olivier Martin 2007, State of the Internet & challenges ahead Monoculture of Networking, αρχισε το 1970 και έγινε πασίγνωστη τέλος 1980 DoD ARPANET resource sharing not research in network technology, λίγες εφαρμογές, γρήγορη επιτυχία, χαμηλό κόστος χρήσης, μέχρι του φούσκα του 2000 CYCLADES experiment 1972, packet switching // telephone network
TCP/IP problems area (2) Τέλος 80' ST-II RC 1190 streaming protocol put pressure on Best Effort quality of service, solution = extra capacity No security mechanisms, myth No support for resiliency, myth IMP ID addressing kept in Ipv6 Not like CYCLADES, DECNET, XNS, OSI addressing --- μοιάζει με κάθε σπίτι να έχει για δ/ση την σειρά κατασκευής του IETF Ipv6: εμπορική επιτυχία = τεχνική
TCP/IP problems area (3) Backbone routes – smartgrid peril IETF LISP, Loc/ID separation protocol ignored IPv7 (1992) TCP/IP split 1978 concept of layers from Dijkstra THE Operating system 1970 TCP, UDP repeat policy/mechanism TCP numbers & acks bytes not packets OSI Layer OK in 1985 – (3:subnetwork independant, dependent, convergence)
TCP/IP problems area (4) Internetwork only < 1983 = Flag Day, IP over NCP NAT break (only broken architectures) (role 1 = address, role 2 = security) IP address + port number = connection ID like address mem in JAVA program IMS αναμονή για πάντα....
CLEAN-SLATE Recursive InterNetwork Architecture (RINA) John Day reporter for OSI reference model Upper layers = one layer (1983) OSI programme collapse, implementers used different protocol machines TCP/IP simple layer model wins Practical IP community BUT fixed protocol stacks = problem also Why? Protocol functions alternate up the stack (error cntl, cntrl – DTP & DTCP), same in X.25 L2,3 same in IP, in TCP in wireless, in applications.
CLEAN-SLATE Recursive InterNetwork Architecture (2) 1 st principle: LAYERS RECURSE !!! το ίδιο πρωτόκολο μπορεί να χρησιμοποιηθεί οπουδήποτε στην protocol stack, γίνεται encapsulated σε μιά άλλη στιγμή του εαυτού του άρα δεν χρειάζεται κάποιος αριθμός στρωμμάτων (layers) LAYERS isnot PROTOCOLS Recursive layer has DTP, DTCP, Management Protocol – F.Goldstein,J.Day pages 10,11 diagrams Moving beyond TCP/IP
Τέλος 1ου Μέρους Ευχαριστώ !