You are on page 1of 9
NIC 15717 RFC 500 AIAA Paper No. 73-417 THE INTEGRATION OF DATA MANAGEMENT SYSTEMS ON A COMPUTER NETWORK by A. SHOSHANI and I, SPIEGLER System Development Corporation Santa Monica, California AINA CoOMpUOF LISMORK euetamne GQnOrenea WD Ououude GORISFENES ne ALABAMA / APRIL 16-18, 1973 First publication rights reserved by American Institute of Aeronautics and Astronautics. 1290 Avenue of the Americas, New York, N.Y. 10019. Abstracts may be published without permission if credit is given to author and to AIAA. (Price: ALAA Member $1.50. Nonmember $2.00). Note: This paper available at AIAA New York office for six months: thereafter, photoprint copies are available at photocopy prices from AIAA Library, 750 3rd Avenue, New York, New York 10017 ‘THE INTEGRATION OF DATA MASACHENT SYSTINS OH A COHBUTER NETHORK® A. Shoshant, T. Splerter systen Develooent Corcoration Eente Yonteay Californta In thts poper we dtscuse aa approach to foten grating daca manaperent systens on a computer Beta Sect fee ene purgese of daca thartog. Our approach Xe dntepration suggests the use of 4 comon tan fiage and trovsistion interfaces to target da Senonenent Langurges. Properties of the common Tanguage sve explored, snd a method of implenenting Thevtranslation incerfaces sya neta-eompiler is Gesctibed. tire flexisility eam be achieved by the UeeGrratnacural Lannwape processor that permits, Ghee requeets to be seated da Pnglish an@erans— Yared tate the foraally structured comon language. Hinaliy, sere conclusions. are craw regarding the desteabitiey and feasibeiiey of thie approach. ntrodu ‘The need for date sharing arises in many sppltea ‘on arenes Fyamples. are reservation systens, hospital rotworse, Library networks, centralized oF Ueteibueed boncing evsrune, and niideary ond govern Rent information aystens. these eystens seewlate Targe amounts of data chat are of Interest £0 = Mise community of users, The Feoulting data bases ay becone tee Latge for econesteal duplication for Ufferene applications. Aside fron that, usiog the Sine daa for ditierent applications often requires Siectortng Jocor a cask thar ds ctfTicute and Erpeneiver, Finally, there 1s 0 groving need for Febictine aad on-iive data shoring, Computer net Sona provide the basic facklsehes necessary co eet these needs. xaaples of conputer petvorks are éiscussed 19 the Literature ls 2 3 8. For our purpose, 1 Tecides'to know rae a computer network ie a Colm Section of computer systena fntsreomected through na previous papce? we discussed a number of approaches to data ehartng Se computer necworks ser Gifrceen appronches (labeled a6 the central creat Mhtesrstets data transformation, and stand Stlecd approaches) have properties thee nake then Stlaceive and applicable under different conditions, Tie chose to pursue our investigation of the fnte~ rated approseh because it has the advantage of formttring the sharing of daca snong network nodes Eftnout.tacerrupting. the conténued use of existing, Be hgagencat syetens and octaeing data bases: ‘the tnteprated approach supgests the use of = coonen Language snd teamelation ineerfacer to car fer dace nonoresent Ianeuaqees Therefore, 1€ per Bite ehe continued access of a particular data Bisigencoe spares either through tes local dats nantbenent longuage se throush the como Tanguane. Neucvers the uve af the connsn Languaze permits the acceas of other date managenent systens #9 the net~ SSE ha tadteton to the local one. Ths dval aode SP bperatton would alley users of existing data seeeekehe ayecens covse the, conmen Language factl~ Tey without relinquishing these previous. local node [Properattons nis da.an Seporcane advancege of che seecbrured approach, since it allovs data staring € fevintrotuce’ inte Computer networks im an evolU" Cfonuty meaner ond fa nore 1ikely to be accepted by CaeETig ances. Anceher advantage of the sneesrated SRIESAGh io'thae te pemmsta the covextstence of nay Syutensy thus allowing for further development of Se gyatens and their fntegracion into the netvork Uithour disturbing extscinn ayscens. eee, ve prosent an overview of the intarrates approsch, then diecuss ehe propertios of the conon Tnneuere’ and cur expertence é9 implenenting eens Tition foterteccs wing a mete-conollers overvtey of the Intedeated Approach, conceptustly, the Intesrated aporoach provides « vor in which sti unseen a computer netvork con Mecoas sil data nanapenent systons, a= shown $0 Figure 1 Auuser, regardless of bis physical Toc {Elons can Aooue a request expressed in the coon0n Snambape to be processed by 8 remote d2¢a nanaxesent Zyeten (OMS)s. an inverface wodule associated vith Tha IMs seconts the requere and transtaces 48 Sato Ehe‘ianuage of the iiss the Tes porforas the Function requested snd getueas che Feply t9 the vser ove thie all dacs wanstenent Cunectoas are performed Bylexioting deta mananenone aystens. The Cask of Gtithe request. there ia-ene Interface for every Giea managenent syaten.. The returned reply can 6ls0 Selteenatormed inte. s coman Tanguage Corcat. Ts fact, thts fe a usefol concept 1 ve wish to” conbine Fetufned dota from nore han one 95 into e sirale Tepive nie eftuntion arfaes when data are dts Exibuced on several systens and need to be shared for s common reply. An exemple could be & hospital ARivocks where overy node fa's Local system anso~ PESteD Caen a hospitaly ve mgpne wane to tnceprete Gnforsatton about ali local blood banks so thst they could be accessed Jointly Dderendtan on the physical location of the fater= faces, tuo main variations of the Interrated Epproseh ere poosttles In ehe intograted-atstrtbuted sBeiaeions the interfaces are physteatly co-located Miun the corresponding ata managenent systems. as Shiota‘ in Pipure 2 in the Saeesrated-cencral. vari Selon, all. the interfaces are physically located in SReRbae and inplenented om that central system, as SESS Figure 30 One uifference in properties Sotucen the teo variations fs that the fotesrated- Uisteinuced ene hes failesofe charactertaties, be ‘use the failure of one sode does not disturb data FRISTEGIAN TS TyPirted by the Advonced Research Projects Agency of the Department of Defense wader onceact Noy DACLSH 3-0-0080. gsr 4it ‘accessing between other nodes Breted-central approach shou fEent, becnuse all tnterfaces a ‘Belngaiged in one central systen. However, the inte- seplenented and ‘he toterfaces descetbed above are arsoctated with the serving node. An additional variation ts To associate an interface sith the wedng node, The Using dntetface can perfors eddicional Tunetions or Gone of the functions of the serving {aterface: thus, he nature of meseagee between nodes Will Vary. For eranple, {can sceept a request Fron the user fn che conson Innguage and fepresent fe n cone Gncemediate languape before sendlog Le to the Serving interfaces This will permit more compact descriptions of requests co be sent between nodes: Gn the next section (on the common Language) we dis~ use an snample of a serving saterface=-a natural Usnguage processor which teanslates English requests inte a precise intermediate Languase ‘bn Soportane problem asroctated vith che inte= grated approach fo the conerol of eratfity” [efor~ Exelon about the source and destinseion of requests Gna vecurned data must be coordinared, Ths fune~ Choa is onually taken care of by the message svttch« ng mechaniams of the computer network. The Situs tits is much more complex in the cage of data Seturned fons Glatrituced data base vifsh need to bbelinteprated. Another protien thac cannot be over Booked fe that of privacy and security, Te appears thee a cootral pode uhieh will check the Legality of requests according to a central authorization fine vould be an appropeiace end efficient vay to solve thia problem ‘The Common Language The use of cmon Language should disengege the indivigesl users from concern as to where the Guta are locsted and Sp what for daca can be Sccecued or stored, Sone obvious properties of a ‘eli-datined cones language are that tt be easy fo ose end learn and that {tbe capable of express” bg the functions deatred. In addfeion, © eomon Tangoage should te bared. oo gensesl enough data Etiuctures 0 allov the reference of coxplex Unfornation, Currentiy, se lini ourselves to Merarchical dats strocturee which ae povsefal though for a large range of applications. (hierar~ Chicel structures are discusted Inter in the sect Lon Shout the fatemediate Langue \\ ae inure 1. cons [ga |_| se : lll Tawmarer | saStider intual Representation ‘The design of a user-level Language usually sqvelves s conpronise between the properties Honed above-vease of uso and Functional expressive pouer. It aay be advantageous to separate these Characteriseice-sthet is, have 4 uaer-orlented Lan Guage for case of use, and. forma! totermediate FEnguage powerfol enosgh te xpeets any functions decised, "the dntermediace languepe can be ss com plex es one Likes, since st ts invisible to the leer. A transformation procens se necessary {¥o9 the user-oraented Janguage £0 the foremediate Leo punee. In the next section we deseribe in detail four Savest patton of the functtonal properties of the intemediate Langeago, which we esl! ILDS (I~ fermediate Language for Duta Sharing). Because TLDS fra precise Ianguase, it {2 possible to build translator interfaces fron ie te earget data menager ‘tent Isnguazes by uring a nets-cenpiler. Our Sparfence tith inplenent tog tranaiecors 15 decertbed Avvser-orfenced languare that fs most natural, of course, is English. We make uae. of COURSES, an Srperinental English-leapuage INS. Placing. the CORVERSE Mfront end,” a natural Language translater, ‘sp sldftional block fo Figure 1, divides. the Mine representing the common Language into dnpot and oucput parts. The input Jean Enelish request on UniGh the CONVERSE front end’ perforne eystex isiyeia snd senancie interpretation £0 produce colludefined intermediate Losguage (IL) Featencate {is ice ourpats Our expertence ith the ODRVERSE. Ghturel Language tranelacor suggested the following shone. (represented by a target daca base) requires Cranslation process Which 1s nore pover‘ul thor th Syaraxcbased parsing techniques on which COMVERSE Bobaseds A foraneic concepeualfeation of he asta base to required thich repreneete the porssble English requests addressed to the data bese. These Conceptey which represent. the poseible rer view of the data’ base, seed to be represented ce the machine Waser requests are to be successfully Interpreted. Mich of Our vork 19 currently ertented in this direction. Aunatural language processor also needs fo have niceas to information about he deta base in onder to "understand the Eaglish request. There~ fore, with every data base, 2 data base description (G00) Ie requized vinich concaine the sexantic cor~ Cente of the data base, This suggests the use of a be'shared on the network. q3-as7 Hgure 2, The Integeated-Dfetrtbuted Approach c entta WO06 Figure 3. The Integeated-Contral Approach > rare Note that if a precise user-ortenced language 42 sed, to Information shovt the data base ia neces™ ry, llwever, a precise knovledge of the data base te required by the user (which fe 2 comon Sssonption made when using current date mansyenent Tanguases Functional Properties of the Intermediate Languoge Our natn concer tn the development of TLDS sas to Wentify the possible functions vnich operate fon dara structures, Ta putting those functions Soro a eystax form, we attenpe to present. the func {logs clearly even at the expense Sf having a Leos Concise fora: in vhat follovsy the Basic nesunpe lone about date structures aro presented an a detailed description of the functions operating on then {5 given As nentioned earlier, we decided to Lait our- selves to hierarchical data structures because they are oufficiene for s large minber of application teas. "(In the future, we miekt expand to nore yneral netvors data steuctures). Hierarchical data Structures (sonetines called tree stractures)y veh fre comon to most exleting data management systems, Fepresent a logical way of orpantring data, Beseri~ Eons of Merarchical data structures ean ve found fn pany references (for weample. J+ 8 purpose of otroductng our eerninology, Ler ue in Figure t. Ihe nare of the hierarchy ta RD (for he Resesreh and Dovelopment Divieien), it has teo branches, roprescnted in the leftmost table te Figure 4: "Every coluan in that table, except PROJECIS, 42 celled a "data elenent™ (OE). The leneats saa coluag are called "vaduee" of the DE. ‘Tovey "TECH" apd "DHS" are-valuos of the DE. "ENAME" oes in that table are called entries, (ln Pagure areneries are numbered for the purpose of later reference in exanplos). The coluan called "PROJECTS fs Called "ropeatiog group" (8c) because it hae suasy instances of projects sasoctated with every the purpose of unified references ing eetty. Fe “tnancnes™ ts size considered to he = rope proup, Thus, im this exemple, ue have hres repeating groups (BRANCHES, PROJECTS, IOYBERS). Each cable, hovever, 19 an Snstance of a repect tng roup. Thus, the SRAICHES RE hae one. tastances, the PROIECTS NG hae two dnstances, and the MEMBERS RE has {ive instances. We say thet three Do and one Fepoating group “bolong” to the DRANCHES RC (hi BHEAD, BEE and PROJECTS Fespectively). Similar Kelationships hold for ehe other Tepeating groups: Note that, fp gonerai, a repeating geoup can have tore than one Yepeating group “belonging” to ft~ For the purpose of sinplicity, ve aasune that DE ames and RE. nanes are unique, even hough uniquenese Of DE nanos tn aifferent Rea is net necessary. aimee the Des ean be distinguished by the Re they belong ‘The tems "Merarchy", "repesting group" (R6), aata etenent™ (Dt), "valve", and nentty™ ctscuased above Will be ured ia the follovinn desertotion of he tnternedtate language IDs." First, ve describe falkernel verofen of ILDS whose syntax de given i he appendix. “Then, we explain aidsttoral possible functions vhich can he added to the Kernel veretons depending on what level of omplenity one chocses te employ. ‘the Kerno! version hae four functions: creates delete, update, and query. These functions a eseriied in tenes of theee basis clesente? quell Flor, output, and replacement. ‘the qualifies cons fate of functions’ thet can be appiisd recursively to's hierarchy co determine uhat entries ually. ‘The enceies that quality can be operated upon by 4an output function to produce values, or $y a f= Blacenene function to chanae values in those entree. ius, the create function has only a roplocenent part, sich spcsiftes hev to create nev entrien} the Solute function hee only 2 qualitice party which elects encries to delete: the update Function ha S'quolifter pare which selects entries to be nod! Esou according toa replacement part} ana the avery function has a queléfter part which telscts enteiee fon which an output pare operates to return values Im the kernel verelon, the qualifier part 4s nost powerful, since st te composed of recursive applica Hons of elenentary "entry-functione” (efunction ey aaa sn the eyntad. quatitter = efunetton (qualifier) neana that 4 qualifier 4s formed by applying an Thery-fonction on entries thee qualified by the SELEY quatitior to get a new set of qualified Eteetegs Thos, an entry-funetion 18 one that naps a of entries into sete of entries. An addi Clonal centriction ena qualifier is chat It repre SSUES tee of enertes of the sane repeating grou. Ihe tact fe reflected in the structure of entey~ Farctfons, chich will be examined Later. To ter Sinate the recuretve form ue need, of course, & forminal qualifier {equslifier) , hich in our ease ava becarchy name (imme). Thus, the frat Geiyttunctton teva quatifter operates om the named Merevenys, the wee of a béerarehy nane ip a quali~ Fler slay piovides for operations across multiple IMerarenien co be expressed ‘the syntax for: very entey-function bas on "eptied” RO aceoed= ved ite fey ehich ia the RE (or which enttice SeStsty arcet the application of the entry-funet ton, HSIN Mesteote’ehe tnptiea RG of every entry faction vhen discunstog each entry=funetion. Flsets nevever, ve need to explain sone adaittonal Teros'uned in the syetee., Tue term fval" nosne a SSlus Se°s sce'of valuce which can be elther explie~ Yelp specttiee or extracted. fron the data base (cet cepintn ister how values con be extrocted). ‘fhe term col" Ceoiven) is either a DE ors recne— Tove appttcation ef erithnetic operaetons on one of foce Diss ‘Tor exsepiey "C(SALART/12) +30)" Sa 8 BepaTteaae” Ganteh bighe sean. monthty eolary + $50) eens teil predicate) nana one of he come Doniy used predicater: equaly greater, Losey Greater or evals Lose ov equal, not equals The Rona feop" (Gor cpesater) scandy For one of the Shevaterd on o cet of velures any averages maxinun GEicer wietmun valve, sod count noxber of values Sn the bets rhe significance of the entey-functions ts that Guilstying entries inva Mevarchy. The entry Finetione are described below! © Select, The select function ts the most com non quatitication function. Tes form fe electecol pred val, whieh caans: compare according "query valve specified sy "eel" with the fait side iE the comparizen succes, qual ing entry. £0 the RG "inplied™ by the DE tn Need’, For exanpte! (SALARY EQ” 12000(KD)) theme! for the go eivision to compare each value Gnehe "sALARY" "col (sere Bef de a DE) to 12000, na 18 choy oro equal, qualtsy the entry in the RE NENGERS", which, £8-te 8G tnpliod by the DE SALARY" (che "SALARY" DE belongs co” the. "MEXBERS" Bo); Tiusy in Figure 4, catries 10 and 19 sn the TMRDERS” fe quolify. "An example of a "eol” which fenot a sinpie Of far ((GALARY/12) EQ. 1000 (RD)), hich qualicies enertes Scho "IENERS" 5G wish fonehiy salary equal to 51000. This, of course, © Group, The group function operates across aevele'of fie Herareny. Te qualifies entries. in a°'Ro according to 1 R04 operation (SOP) on values Jn 2 no siten belonce to ft+ For example? quality Elerion tm the "PmoJEGrS® RO°for which the averare SAAR" per peoject ir greater than 15000.” Tex fom ts: group = rp sop col pred val. For the exeaple above ve have: (PROJECTS AVG SALARY GR 15000 (@0)) (entries 5,7 qualified). The inplted #6 for Sie proup Function fay of course, the RE which opente in the function. Wote that the group fune~ Uidq can operate across more than one level, such fe the entries in the "BRANCHES" RG for ubich aver= Spe “SALARY” por branch 1s greater than 15000. Scope. Tats function, too, operates across testis Sf'tlie hievarchy.. Ive purpoze 4a to. qualify eneries ina Re in one Level ses result of entries Ghee guslified ine RE So another level. Yor: Chanples suppose chat we vane to qualify branches 4 they contain at lease one project that quslitied #9 our previous example: We apply & scope function Ghuteh Te sicply-cscove © tr). to the previous exam flees follous: (@RNCHES. (PROJECTS AVE SALARY CR Ys000"(RD))) Centsy 2 qualified). e can elther Mgeope up or "scope doun", depending on whether the Seope RG (Res) ig at # level Bigner ce lover than the inpiied Re of the qualifier te operates on (RC). then, Necoping up" we quality entries in Ro, for Untch there de ae Jesse one. entry qualifsed in RC, Sines. "scoping down”, we qualify ol} entries in eg for every entry thee qualified 10 RG. The privious exsgple Lt one of coping up", "tho follow The example fe one of "scoping soon": GENSENS. (PSA EQ 10"(H0))), whitch simply qualifies 8l1 nesbers of the 10 project. + compare. The compare function qualities #5 ently Sertstpartng values fron to columns for chat Entry. For enanpley suppose thar ve wane Co. qualily nenbers that have the game nunber of yeses of Grperience (EEA) and years with che company (Cheats). She eynton teprecentatton for this ample SS! Ceo eveans £0 CYEARS. (80). totrtes cualsty ty comparing, YEARS and CYEARS patcvise by entry Generies 8, 22, 13, 15, 19 qualify). The syntax foo for this function is: ‘conpare = COMP cot ored col. Beth "col" cleventa in ents function must belong £6 the none R¢r This BO a, thereforey the implied te © Loop. This function te sintier in forn to the eonpare fonction: loop *t00P col pred col However Shatesd ot conparing values fron the colums pair= Wises te conpates one value fron the first column Uioneach value from the second coluant then it Tepeats the process for all other values fron the Fifer colunn’ For example, suppose ve want to cual= fey entries in the SBRAICHES™ Re for whieh beach heed" (BHEAD) Le also a project lender. CPLEADER). Ths ‘Skanple se exerested Sy’ (LOOP BUEAD £0 FLEADER(RD)) (entry 2 qusllss)s. Note tne. the colina can belong to ulffecent Roa} therefore, thee order te Inpor= fares ie dope, the convention that the lnplied 80 [2"Ehe one to unich the first colin belongs. 4 Logica! Qualifiestion, This function 4s a conpound one vhich pernite qualtfiers to be Joined Sopveally by the AND end OR Londeal operator Cop) Leds expeesacd in a Fecuraive form Co. allow quali [lets torte joined logically az sary eiimes a6 neces Gnry. For exenpiey if ve wane to qualify mesbers Ghose saieey ie yroater than £18,000 or who have nore then 6 years of experience, we Join t4o select fune~ Elons a6 folleses (OR (SALARY GR 18000) (EYEARS By (Ro) Cengeies 10,15 16 qualify). Tt as peruttred to Join qualifiers chose implied Ris are in gitferent fevels in the hierarchys however, the tnplied AC of the jotmed qualifier f2 always tie lover of the to. W3-417 ten an entry-function 16 being applied to a qualifier, scoping Functions often become rocun= dine as in the following exampler (CeuzaneR oq KENT (PROIECTSCANAYE EQ ON (RD)))) hich neane: in the BMS branch, qualify these fetries In che PROJECTS RO for viich project Leader Is ker. "Cin ehis case only entry 5 quslifies)- snot necessary since tho fora: CPLZADER EQ KENT (BSAE EQ ONS (ROY) iplies 9 scopieg down botuean the evo inplied Ros TEMEERS" (from SAVE) and "PROJECTS (Erom PLEADER) « Scope functions are necersary, however, s2 can be ‘lloseraced in the Eolleving @xanple (anmess (PROJECTS (VEARS GR 6 (20990) wich means: for the project which has a ember With oore than 6 years of experiences qualify a2 Resbers.. After selecting on MYEIRS® “entriea 13, We quality. Afcor scoping doom to "MENGERS" RG, enties 15, 16, 17 qualities. Me mentioned carlton that values can be ex tracted from the data bases vis te dove by. ayply— fog a valucrfunction (sfunction) to a qualifier ‘Thus, the valuesfunction maya entries into values. ‘The simplest value function ie? veunction = col Te allove the extrectton of ali the values under a the example of Figure 4, the form (SSAHE) (SALARY 59'12000"(7o}) returns the Values FONG and TEE. A pore ceuplex value-funetion t= the singlo-value-fone~ ion (av) eich extracts eae value only by appiv= fog a sct-operation (scp) on the values of a column. For example, {JUG SALARY) (EYENRS GR A(RD)) extracts of experience. tn extension of the sel {x the ultipie-veluc-funceion (ov!) which extracts eul- tiple valves by gFouping the eateies st operates fone For exanpley (PROJECTS. AVE. SALARY) (EVEMES CR {Cipy) exeracta the average salary per project for Imenbors uith nore than 4 Years of experience, output, but ie ie also weed eo extract valuee to be sed for further qualification. for example, to ‘Qualify nenbere with salary greater than the aversge Salary, wo need to extract the average salary ueine Svalue-fupetion as follovs: SALARY GR (AVG SALARY (H)). Another output function te TF (enve- fale), which revums false Sf s0 entries quallfy, othersise trues, This function ia seceasacy for (Gr) (PLEADER EQ LOVE (RD)) to represent 2 project leader Mttional Propertses of the Intermediate Language The kernel version of TDS Ss not an attempt £9 represent all possible functions on a hierarchical data base, rather, it provides us with a pover‘ul enough oo) for our experineacation with trans developing Lb3, we recognized many Cunceonat Properties vnteh sre not expretaed dn the kernel Sotvions Ue describe. ther bristly below onde BEES RSE ATP OS one hap mony levels of qualifiers Sahedded in it, one for cvory application of an entey-function. Te night be useful to attach an output function to the Gifterene fevels, rather than to the top-level (qualifier only. "Outpue attachnent is Seportant for Zoparating the output process from the qualifies lon process. ‘onery=funct loa to apply tote. This will form s "eves etructure”™ whose osdes are qualifiers and whose arcs ere entty— Functions. Ovtpat attacknest can be made to the holes of the tree: An exey way co express both This Conctson sed the output ettachnent function de by a numbering schene for the qualifiers in dif Feront levels. Fatey-functions and. oveput-funes tons rofer to the quslifior they operate oa ty its summer. = adaieton B ‘those include she" siporeantfetione of ostpot formatting ane 4 Adgieional reference functions for quslifica- a, “Those Includer vers" function--to qualify, for exanple, projects for veh every monber his salary reater han'815,000.” This function is also eallee "the nivereal quantities", = "sume" funceton=-t0 qualify, for examples pxo- jects with the sine project leat. fron ordered Liste in a data baer For enanples If fist of people fs orderes by age (oldest first) i 43 data basoy one can cbeain the youngest ef then by Toquosténa the ast person on the List. fangeone etme ere in rors Naneope, such a5 conditional statements (IF seatencres) + Fepested execution (FOR statements) and subeouctnes Addicts tata a fanctiass. th the were version of IES, at nate aphasia ie ob sualitceaeton for the purpose of fetticnal af data. onysadesetonal Gata mrgngenene.tusetions ate sos eyed tn the disecstoredelOr tye date anvtens ton functions and dace wodieteation foncelsnee {ihe kernel version tne only a bane, Slope te- incenene part forthe opie’ and eveute FosceZons.) the Losnclator interfaces. Ye decided to taplenent swverat transtator inter faces in order to quia actual eapecienees porafois Tencity ontorescen probiens, sud develsp © settee {a of she fenatoiticy oF the Tnceneatedtopeeee. We describe below tome fusetlons that se teteefane rust porforn. 4 Tranclarton from ehe csamon language eo the “language CML]. The Crane toto be ealTovad To" the spectttc Diy tn'ehat te needs te take mivantage sf spect Funct dons that can be expronsed by the De. Te those are two sve of fenresent thy a rowucst tn the Diy che more efttctent ane must bo chroens Tp. the case OF TLDS, we have to consider whether a qlali= W-4. char vith multiple entey-functions ean be exnressed Soinely through a single request in the TAL. fo appropeiate refusal of a roquest. Tf the tar~ got asthtmnaganene system 1 Hot canable of perform Af 2ehgiece cnpressvd tn the coanon Language, tien ANE ineettace seeds eo refuse te praperiy, fatter thin Tetunine Te as "esaot be done" an indication fete why fe cannor be done might help the veer to Foohrase his question Translation of returned data. Replies froa Coning fron diffurers aystens (Chis potae was s~ SEE Gitiicese “furshtroore, a common data format SS neceasery far returning complex gata such 03 EiPekSSaeca in order eo Se cuepue on zone focal 1/0 ovice needs to be shipped to. another remote aysten ‘eM atene nove difterent 1/0 devices. A "post processor" at the tenote node would then receive The ceport in the conon data fornat and ovtpue $¢ gnan i/o device dostensted by the user who Tasued the requests he prcblen of shipping data and {lex fn computer networks has been recognized a9 & Eescral opgy and vork 4a this direction has bee FeTeedtegTE! “ET fenceal mockantone for data trans For ere available on a canputer networl, then Ue fnterface should use then for returned data Translation of error neneages into a cnsnon verbs Tatnate Sveny tee management syotea ususlly EEPEE CUE crrorsnossage formate Te ie useful to teers nave to te fasiliar with only eae error format. ‘+ control functions. The snterfacel need so hove SSning from ei€ferent uoers on the nereork. TMs thpe at corerol nechaniea anoulé aoraally be part of Petvork functions peeformed by every node. "Ai Uiditions! cencrol meshanion iz necessary in order teltgtabiish cominieation between the interface end the target data manoperent systen. Control. fune~ Clon sheutd use interprocere comunteation poche Mant aveltabie on the conpucer se4vork, such ae the caperinentsl ene dererizes inl ur experinents vith inplenenting incerfacos con contrated in the area of erenaLetion Fron the com SGatyinguoge te the carget data masagenent. Tanguazes: Because ve wanted co lesen gore about this process Sinciuston ves that. if tenporary seorape functions for thete equivalent) are availabe in the target data pancgesens systeny then it 43 possible toon $akge"ane apparent functional pover of that system by splitting conplee cequeaes 4a the eonnon Tanuane Theo's series of requests in the DML. For exseple, [nthe cae of DS/2 (an SDC dats manogenent ersten) SPoGhGsEE quattt sor" suneeion reseefets the daca Base for the nent poration according £0 the qualt- Fler. The tranelacnr £9 98/2 ean take adeantane Of this fonetion be represent ings say. a complex Toltoves be a sispe request that can be expressed fn the OS/> IM se penttened fore, we tnplenented the exons racloe ty apes ob aesexcconplior (on S80 product raion Negi “fuletasciees wring prostane She'Shee Logeuanes a tank thae wie quiee complex Soe SEISLCRMERE' al Etplobented tue translators tzon an earlier version of TLDS (hich was based on felneional data structures) to the 15/2 and TKS) Eltzcanagenene langusges, Be spent 2-3 tan-nonthe To'feplonene and Gebup each Eransiator on the GHC SIMBAD CONCLUSIONS ‘the integration of data msnegencnt ayscens on & ‘conpter netsoth night be achieved by uring the ethegelony of 8 comen Language and) traneLation Uneeetaces fp addteion eo the advantage of fnte~ tracing extsting date nanogenent eyacens and da PEI approach facilieaces the evoluttonary develop ent of new aystens. The dual mode of accessing Gera-casing the daca managenent Lengusge or he SStnon Tanguace-rshould be of geeat value for ERUGtr abapestion to conputer network dats sharing. XGiendsancage of thie approach is that Ceanslator fotertaces sé o level of Language translation. Hovevers our experience sbovs that the ¢ranelation ee dacs wenagenent. languages 15 a manazeable task, Firthersores sassidering that conteol, eomuntea- ions aad dota teaneier functions necd bo performed [loGly anpeesch for deta sharing on a netsork, the ldttion of eranstetton taske may not be so siantfi~ the weparation of the cormon lagguage into (vo sevele-vthe tnglish Lenswase for user coneentence UesUy dreenediace Languare for proctse expressive powers-helped us to concentrate on exploring the Fonceional propesties desirable n 8 cenmon Languare. he cooaon anpuane should feprecent @ “union” of Geta monapenent Lengeages, go that tt con repre, SEXE "Uieeses fer al) data management systems. Ost Spproneh uae to identify the elenentary funet ions wePTGateal dete serustures, and allow for their Tecurstve application. A kernel verston of this Teomuage vas defined and experimental translators [roe it eo tuo example data menagenent systens Were fulte by the use of a metecconptter ur expertence shows that those translators were solaeively easy te saplenene with a weta-compiler {about 2-9 mancnonth per translater). Another [eportant conclusion was that 1 i2 possible to ada Emctionat poer to earget data manopencat syste Sy eranatoclog a single request in the common 1an~ Mage Inte a aerser of requests in the target data Aansgemene Language. she use of Taphish ag a commen language {5 not ‘useful For sonbistlented waero. TE is very dif HEGie, “tor exenpley to oxpeess in English = pre~ Else Egpore ecqueses Inveaaietony a descripeton of fhe earget dats base must bo avaiiable £o the Teqosesey whether oF noe’ an Enaiteh input mode 4s Gvailsbie. none syacons ona computer network using Gechniques Ueteribes tn ehis paper offers many advantages {e'Sfoutd Ieod to shoring of existing and new daca, fo praiual acceptance of che computer sezvork data’ Shaving concepts covehe natural survival of better ei Mhnacnent eystens and elimination of Less Siohone syoten gevelopment.. Tes Feasibiiiey is TEGGiueed om a well Setined commuter neuwork with T3417 4 Yaa ced interprocess communication and data tras fer capabilities Acknoutedanent We would like to thonk Gunther Luther of $b¢ for his help and sany Ideas in the development of TLDS" Appendie: Kure) Version of Lps The syntax for ILDS 4s expressed in EXF om except for braces (('}) which are used to tneteate that’ the elenents botieen then sey appear one oF bore nes. Note that parentheses are Cermizal Symbols inthe syntae. (creator)! (cdetete>)! (updater) 2 Icequory>) est extant (ereplacenent>) asieve deLeTE (qualifier) update pate: Cceplacenent>)() auery Reruns (jsouepue>)) (qualifiers) fuolifier #5 ) eunetion i+ esclests| | ceompere> {Sloop>i select ccolsepred eves erour Stpecanpeccole prodvevals Joop. oor seolsepreds‘igva) | (équatstter) va (eofunetfon> (gual igser))I lepvelD Wfonetion t+ sents| Ceavi) I Canis) me ‘eeopo coals |) me ergzcaops) | ( chops ) ‘eeol>bepsenl>) | Cevaloebopsecci™) pvat = ceering>|enunter> Equaliftersi= chnanes Eeplaconents:= €()} n Top aso | oF peed 2} GR | 1s | 6g 1 10 1 x é bop bE beret son = S07 ave | tax Vee | comer 8. a i> ddentifter To shorten the syntax presentation here, the elenents string, {centitier and nimber are not fxpressed tn detail. (Thott interpretation ts the 10, hotmal ones mumbor-vony integer, fixed point oF Hloattop pete nurber! string=-a concatenation of ‘one or more letters, digite and spacial characters, Sdentiticr-vis a seting whcee féere character fa 11 a letter). Aa example of ‘the Foltoving 18 a query veteten 4m che TDS language and the Feprezentatton of ts parsed exces ‘he eucey requests hates of caplovecr in the TIS branch whose salary 2 greater than the amount of 13. £10000 plus 51000 For every year of experience, query 12. egromre0m sae ot “ “epee ep fine sien Farber, David J. "Wetvorks: An Introduetton", Datanalion, Vol. 18, € April 1972, ppe 3é-c0 elt, G. Ce et al. "computer Networks", Computer, Septenbosdctsher 1970" ppe B32. Roberts, Ls 6 Computer Ncewort' munteat 105 Metworks, edited by Abramson and Ku, Preneice sli, 192, Mekey, D. By and Karpy Ds P. Tay Research ‘Conputer Setene Computer Networt™, Department TBM Rasearen Report C3421, Suly 1971+ Shoshant, sy "bata sharing m Computer Networks", WESCOM Confezence, Conputer Networks, Session, Sepe. 1972 (Also, ARPA Ni Document 12623, Stanford Research Institute, Ost. 1972) Kelloggs Cuy et aly “rhe CONVERSE Natura? Lengusec Data Manapenone Systent” Current Statue ané Plena", Proveedlage of the Svrgootue on Information stSenge ad Stty of Maryiaas, 1971. Bleser, Robert E., "Treating Wierarchical Dat Structires In the'spC Tne-snared Daca Manages went systen (FUNS)", Proceeding of AD National Mocking 1967, pp. Gi-Se Eqgles, Robert W., "A Tutorial on DataBase Organization", Kntual Review of Avtoates Programing, Yol. Ty Fare 1y 1972, pps 1-66. Fey, 3. P. et aly “hata Hanagenent Systens Survey MIF=329, Tha MITHE Corporation, May 1969. Feature Analysis of Generalized bata Base wanapemone Systane”, CUMRSYL Systems Comittee, agit DIL. Crocker, 5. D. er al, "Punction-orsented Proto- cole for the ARPA Computer Network", Procesdinge Sfuasies Joye sicc, gps 27-200 Care, 5.4 Crocker, 8.5 Conf, Ys "BOST-HOST Couminicatton Protocol, $n the ARPA Necwerk" Proceedings of 4610S 1970 Sec. pp. 589-908. Book, Buy Sehorre, D. Ve, and Shermany $+ Jey the’ OWIG/360.Syeten, A Comptier for tediing and Inplonent ng Conptlers™, ACH STOPLAN Notices, Vols Sy No. 6, June 1970, ppr 11-29. @3-4i7

You might also like