n ltercs lter-boundedcsn ltercs
Thisstatesthatgiveninsuf cienttime,theanytimealgorithmalwayscomputesonlyastrictsubsetoftheclassicalalgorithm.
UseofKIV:BothpropertieswereproveninKIVwiththefollowingstatistics:the rstpropertywasprovenin14steps,ofwhich8wereautomatic;thesecondpropertywasprovenin45steps,ofwhich28wereautomatic.
PropertiessuchastheseguaranteethatthePSMdoesindeedbehaveinadesirableanytimefashion,graduallyapproachingtheidealcompetencewhenrun-timeincreases.TheaboveresultsshowthatitispossibletouseDynamicLogictobothspecifyandimplementsuchanytimebehaviour,andtoprovetherequiredpropertieswithinthislogic.
Noticethatallofthesepropertiesareformulatedintermsofthedeclarativecom-petenceoftheanytimePSM(thefunction lter-bounded,speci edinaxioms(3)–(6)).Sincewehaveproventhecorrectnessoftheoperationalisationfilter-bounded#withrespecttothiscompetence,allofthesepropertiesarealsoguaranteedfortheoperationalbehaviour.
3.4GeneralapproachtospecifyinganytimePSMs
Inthissubsectionwewillsuggestamoregeneralcharacterizationofprogramswithaboundontheircomputationtime.Ifwelookatthe4axiomsfromthe lter-boundedspeci cationwecan ndthefollowingunderlyinggeneralconditions:
–
–
–
–axiom(3):startcondition,axiom(4):growthdirection,axiom(5):growthrate,axiom(6):endcondition.

