Topic: local vs. global

topics > computer science > Group: distributed systems

absolute vs. relative names
hierarchical structures
event time
global declarations and variables
local declaration of data
localized understanding
Newtonian physics
Petri net
restricted use of global variables


State may be locally or globally defined. State is globally defined if it has the same meaning and behavior everywhere. It is locally defined if it has a consistent meaning within some locale.

In a distributed system, most properties are locally defined for efficient access. Concurrent processes may not have a global state at all. Global, directory information may use replicated servers.

Bulk data, configuration, requirements, proofs, type information, and class properties are largely global.

Dynamic scope (e.g., environment variables) combines aspects of global and local scope.

Personal, economic decisions depend on local knowledge. For example, deciding on dinner requires knowing the available options. (cbb 8/06)

Subtopic: prefer local state up

Quote: in an open system, control must be distributed so that local decisions are made where needed; uses local information [»hewiC7_1986]
Quote: in distributed computations prefer local state, time and names to global state, time and names [»hewiC8_1977]

Subtopic: no system state or global properties up

Quote: the main problem with a decentralized system is a lack of global, centralized knowledge; e.g., where is a file, what is the best route [»grayJN6_1986]
Quote: with concurrency, do not have the notion of system state; might ask how far out-of-step can things be [»holtAW_1970]
Quote: Petri net theory rejects all global properties [»petrCA1_1966]
Quote: in process paradigm there is no global state or global time due to concurrent activity [»stroR10_1986]
Quote: in the process paradigm no system-wide knowledge, no superuser, no centralized management; everything is distributed [»stroR10_1986]
Quote: .NET's common language specification (CLS) is a subset of its type system; unique, Unicode names, no static and instance methods or fields, properties, events, predefined types, closed

Subtopic: requirements are global up

Quote: requirements and proofs are inherently global; a changed requirement may require redoing most of the proofs; verification gets dropped [»berrDM10_2002]

Subtopic: local knowledge up

Quote: instead of globally consistent economic information, have dispersed bits of incomplete, frequently contradictory information [»hayeFA9_1945]
Quote: very important, unscientific knowledge of particular circumstances of time and place [»hayeFA9_1945]
Quote: every individual has unique information about particular jobs, people, local conditions, and special circumstances
Quote: economic change not captured by statistical aggregates; constant, deliberate adjustments, every day [»hayeFA9_1945]
Quote: localized, economic change requires a decentralized solution; decisions must be local [»hayeFA9_1945]

Subtopic: most data global up

Quote: most data (by amount) is global for SpecInt95 and embedded programs [»engbJ5_1999]

Subtopic: object-oriented classes as globals up

Note: classes provide a way to program with global state that is localized to its use [»cbb_1990, OK]
Quote: there is no semantic difference between class and instance messages; distinction is for organization [»xlrg8_1981]
Quote: static CLR fields and methods are global to the type [»hamiJ2_2003]

Subtopic: dynamic scope for configuration up

Quote: use dynamic scope for configuration of customizable environments such as GUI packages; avoids configuration objects that are passed to every method [»hansDR6_2001]
Quote: implement dynamic variables via set and use statements that allocate from the stack; 'use' searches the 'set' variables for a match [»hansDR6_2001]

Subtopic: decentralized globals up

Quote: Clearinghouse assumes one global database that is decentralized [»oppeDC7_1983]

Subtopic: global system state up

Quote: automata theory and classical physics use a global, system state determined by the global, real time [»petrCA1_1966]
Quote: Newtonian physics assumes the existence of a real universe; relativity and quantum mechanics are based on observability [»wegnP10_1986]

Subtopic: local effects up

Quote: in physics, action has local effects (e.g., inverse square law); in computation, a tiny program can clear all of memory [»hillWD_1985]

Subtopic: denoting phrase up

Quote: a denoting phrase can be a primary or secondary occurrence (global or local); handles denoting phrases that denote nothing [»russB_1956]

Subtopic: local names for global objects up

Quote: besides a global name for an object, allow users to define local names for the object [»watsRW_1981]

Subtopic: naming of globals up

Quote: an overloaded or hidden name can cause difficult errors; minimize by avoiding names like 'i' for globals or variables in large functions [»stroB_1991]

Subtopic: avoid overriding globals up

Quote: Java does not allow a local name to override a global name; an earlier version with name hiding caused accidental errors [»goslJ6_1997]

Subtopic: von Neumann bottleneck up

Quote: languages should avoid the 'von Neumann' model of a global memory manipulated by sequential execution; bottleneck between control unit and memory

Related Topics up

Topic: absolute vs. relative names (12 items)
Topic: hierarchical structures (14 items)
Topic: event time (45 items)
Topic: global declarations and variables (33 items)
Topic: local declaration of data (11 items)
Topic: localized understanding (43 items)
Topic: Newtonian physics (79 items)
Topic: Petri net (44 items)
Topic: restricted use of global variables
(22 items)

Updated barberCB 8/04
Copyright © 2002-2008 by C. Bradford Barber. All rights reserved.
Thesa is a trademark of C. Bradford Barber.