BSO - BROAD SYSTEM OF ORDERING

Page is under development and text is still in the process of preparation
(Last change29 August 2002)!

The BSO Manual

Excerpts from the book:

The BSO Manual : the development, rationale and use of the Broad System of Ordering by Eric Coates, Geoffrey Lloyd and Design Simandl -- published in 1979 by FID.

CHAPTER 1: BSO - ORIGIN AND PRE-DEVELOPMENT PHASE

Subject indication for an information network

'Subject indication' is the phrase used in this manual to refer to those facilities of an information system which enable it to be interrogated by queries which have a subject as their point of departure. The user supplies to the system the name of a subject with the aim of extracting information on that subject from the system's store. Frequently the system contains in its store, not the information ultimately required, but records of the names and addresses for documents in which the information is likely to be found. In such a case subject indication has as its aim the identification of documents carrying the required information. The tools or languages of subject indication include indexing languages, classification systems, controlled term or keyword lists and thesauri.

The Broad System of Ordering is such a subject indication language, or, more specifically, a classification system, developed for a proposed world-wide information network covering the whole field of knowledge. At first sight there appears to be a little reason for supposing that a subject indication language for a network should be fundamentally different from a subject information language for information system generally, and it is arguable that the schedules of BSO bear this out. However, there are areas of such uncertainty surrounding subject indication languages, that it would have been rash indeed not to have put the matter to the test by information requirements in view. It will be seen that the result of the exercise bears a family resemblance to some of the document classifications which have preceded it, despite the fact that during the exercise no reference or recourse was made to literary or documentary warrant in the direct sense. Whether the differences between BSO and the document classifications are considered significant or trivial in themselves, they could possibly prove to be essential to the network application.

Origin of BSO

BSO originated in the context of the idea which emerged in the 1960s that consideration should be given to the possibility of a global network of scientific information centres, taking into account particularly the needs of developing countries. The network idea was itself triggered by a technological development, not at that time generally available, but certainly upon the near horizon. This was the possibility of cheap and fast data transmission links. It is notable that thinking about the information network, involving the first steps towards system definition began about a decade before the hardware became generally available. This was an honourable exception to the more usual situation in the mechanisation of information services in which the computer hardware was available well advance of system planning.

The subject indication sub-system of the network was seen as an important part of the whole system. It was vital in such a network that information on the subjects of documentary resources held by any one participating centre should be accessible to all the centres in the network. There were two closely interlocked, but still separable, problems here. The first was that, despite - and perhaps because of - the growth of mechanisation of information services, which in the late 1960s was just getting under way on a substantial scale, a greater amount of subject indication activity depended upon human intuitive skill and know-how than in the pre-mechanisation period ending about 1955. There was, for instance, an unprecedented proliferation of controlled keyword lists and thesauri, for use with mechanised systems, but little sign of common logical rationale in their construction which might otherwise itself be amenable to mechanisation. These human skills produced indexing tools of great diversity for particular subject areas. Did these tools, the detailed construction principles of which were usually not fully communicable, offer a suitable model for the subject indication language of the proposed network?

The second problem arose for the realisation that for often good and sufficient reasons, centres representing the various subject fields would continue to use a variety of subject indication languages, corresponding to a variety of needs. Accordingly, communication through the use of a possible standard indexing language, which all participating centres would use for subject description of documents, was ruled out. On the other hand a solution seemed to lie in a procedure whereby subject information coded in one local indexing language could be converted by clerical means into the codes of another language conveying the same subject information.

Switching indexing languages

It so happened that this solution involving interconnection of individual local indexing languages, by a mediating or switching language had been under study by the Groupe d'Etude sur l'Information scientifique based at Marseilles, since 1963. Unlike the proposed global scientific information network, the system envisaged by GEIS was composed of centres dealing with the same subject discipline, and the particular discipline used as a study sample was the Science of Scientific Information itself. This difference is of some importance when considering the transfer en bloc, of the conclusions of GEIS in their 'Intermediate Lexicon' project to the context of the global scientific information network. A key feature in the GEIS scheme was the 'equivalence' or 'conversion' table in which the code for a given concept as rendered in one indexing language was coupled with the code of the same concept in another indexing language. It was assumed that such coupling was practicable to the extent that both indexing languages were, in fact as well as in pretension, lists of terms each of which corresponded with a definite and unambiguous concept. In fact the term 'indexation' was reserved for the process of concept analysing a document and assigning a code accordingly. (The code could be a notation symbol of a classification, or a descriptor or authorised term drawn from a thesaurus or subject heading list).

For the simplest case of a network in which the participating centres taken in aggregate used only two local indexing languages, all that was required was a pair of 'equivalence tables', one leading from language A to B, and the other from language B to A. If there were more than three indexing languages represented in the network, then it became more economical, in terms of the number of pairs of 'equivalence tables' required, to employ what has been variously termed as switching language, a mediating language, or a communication indexing language. A message (i.e. a subject request, or the answer to a subject request) would thus proceed from centre A (using local indexing language A) via an 'equivalence table' to the switching language, and then outward to a further 'equivalence table' reaching its destination, centre B, coded in the form in which the same subject is rendered in local indexing language B. This system, which is exactly analogous to a telephone network, would require one pair of 'equivalence tables' (one subscriber's line in the telephone analogy) between each centre's local indexing language and the switching language, ; whereas, if there were no switching language employed, each centre would need to construct, and of course maintain, as many pairs of 'equivalence tables' as there were languages in use in the network, minus one. It has been mentioned that such a system was expected to work, subject to the condition that the various local languages were in fact concept controlled. By the same token, the switching language itself would need to be one in which each representation (notation symbol, or term) corresponded to one, and one only, concept; and in which for every concept there was one , and one only preferred representation. The form and arrangement of the switching language could be considered a function of the kind of use for which it was intended. If it were required only for simple matching, its arrangement would be immaterial to those engaged in day-to-day operations of sending and receiving messages. If additionally it were intended to employ such a switching language for hierarchical search, it would be necessary to incorporate the necessary hierarchical linkages into the language. However from the point of view of constructing and updating a switching language under controlled vocabulary conditions, some form of schematic arrangement, on the lines of a classification, is probably mandatory. This schematic arrangement might not, however, be the form in which the language was most conveniently held in a computer store.

It is to be supposed that the idea emanating from GEIS (which was later given quantitative elaboration in a research study carried out at the Polytechnic of North London School of Librarianship) because it was the only one of its kind available, must have affected the thinking of members of the first bodies charged with the task of considering the global science information programme, when they turned to the question of subject indication.

This text is followed by the last section of the chapter Steps towards clarification (pp. 3-8)

CHAPTER 2: APPLICATION OF BSO

The limits of broadness

The first task of the FID/SRC Working Group was that of sharpening the somewhat indefinite terms of the remit entrusted to it. The central question here was to try to decide in the most concrete possible manner what was to be understood by broadness as a feature of the proposed Broad System of Ordering. What should the determining principle be, which would cause some terms to be included in the scheme as sufficiently broad, and others to be rejected on the ground that they were too specialised?

Arithmetical Approach

Several possible approaches to an answer to this question could be foreseen. The answer could be purely arithmetical - a stated total number of terms in the system could be settled in advance. The answer could be based on a particular property of terms in relation to the classification structure yet to be devised - namely the hierarchical level of the term within the structure. Or, it could be based upon some inherent semantic property which a term might or might not possess. Or, yet again, it could be based upon some formal linguistic property of terms. Finally, it might be possibly based upon some sufficiently objective social property or phenomenon associated with the term or with the concept denoted by it. An obvious thought here was that a social property useful as marking cutoff of detail might well be one closely related with the purpose which it was hoped BSO would serve.

The first approach to the problem of defining broadness, or cut-off point, for BSO - the laying down in advance of the total number of terms to be used - had some special attractions in relation to cost predictability, especially in the context of mechanised exchange of information within a network. Clearly the cost of the computer processing of such exchange would fairly closely depend upon the size of the interconnection language to be traversed in the passing of each message. This dependence is probably less significant now (1978) than at the outset of BSO development in 1973. With the expected future use of microprocessor elements as customary computing hardware, it is likely to be even less significant in future. At the beginning of the development of BSO, it was provisionally assumed that the full scheme might contain 2000 terms, and in the first draft submitted for comment in 1975 there were in fact 2100 terms. This draft was faulted both on account of its omissions and on account of alleged over-development of detail. Such criticism on mutually opposed grounds might have been crudely interpreted as a justification for the middle position taken by the BSO draft. However, the tenor of the comments themselves pointed to a great weakness of any solution to the cutoff problem based upon a prescribed maximum number of terms. As the approach to the maximum is reached, the question of what is, or is not, to be included in the system, comes to depend upon refined judgments of the relative importance of candidate subjects. Reliability in such judgments or judgments reflecting a real consensus are hardly to be expected from practitioners in the borderline specialties themselves - these specialties are, after all, often in competition among themselves for social recognition and funding. For this reason any purely arithmetical characterisation of cutoff in terms of the total number of subject-terms which the system is to contain is likely to be unsatisfactory and tendentious.

A brief side-glance at the arithmetical size of the scheme as a result of a cutoff criterion to be described later may be in order here. The 3rd revised draft of BSO (1978) contains about 4000 terms. The 18th edition of DDC, an established general classification for books has about 80,000 terms, so in approximate terms an average 'broad block' of information which can be designated by BSO is 20 times 'broader' than typical information at book level, and 3 times 'broader' than the information units which can be designated by the abridged UDC.

Hierarchical Approach

Many comments received on the earlier drafts of the scheme assumed without question that cutoff could appropriately be defined by reference to some-hierarchical level in the scheme. The arguments against such a basis for setting the limits of detail of the scheme are formidable. It can be contended that the policy on limit of detail, far from being derivative from the exigencies of the structure of the ordering system itself, should be independent of that structure. The structure is for the purpose of ordering, not for delimitation of acceptable detail. Furthermore hierarchical level of a given term is one of the most unstable features of all classifications in face of necessary changes required by the arrival of new knowledge. Much new knowledge arises by the fusion, following the discovery of common properties, of two or more hitherto separate subjects on the same hierarchical level. Whenever this occurs the separate subjects and all other subjects subsumed by them change their hierarchical level. Another consideration is that a statement of a hierarchical level is often made for explanatory or presentational purposes (for example in BSO 212 ENERGY INTERACTIONS & FORMS (ANY STATE OF MATTER). Obviously alternative presentational strategies are possible, and they will to some extent depend on available type variations for display. Also a chosen strategy may at some time have to be modified because of the appearance of a new subject remote from the hierarchical statement in question. Hierarchical levels are thus determined both by logical imperatives and presentational nuances. Both factors are subject to necessary change, and their states at a particular moment should not be the determinants of system detail cutoff Finally the practical importance of a subject is by no means necessarily correlated with its hierarchical level. For instance 923,70 BASQUE LANGUAGE, being a unique member of a set is on the same hierarchical level as 921 INDO-EUROPEAN LANGUAGES.

The lack of agreement between natural languages as to the incidence of 'logies' and 'graphies' probably reflects the fact that mental organisation - the central characteristic of the kind of knowledge which constitutes a discipline - is not an all-or-nothing property. While one can perceive intuitively that, for instance, Chemistry is a more highly organised system of thought than Reprography, this is not to say that Reprography is not a discipline. Indeed, it would be quite hard to identify any subject matter which has generated literature, which can confidently be said to possess zero mental organisation. On the practical plane of handling subjects found in documents, no hard. and fast line can be drawn between disciplines and non-disciplines from the standpoint of mental organisation of the material. It is only a question of more or less, and no simple method of scaling this spectrum of more or less mental organisation was available which would have enabled the FID/SRC Working Party to apply a quantitative criterion for cutoff of detail.

This text is followed by the sections:
Linguistic Approach: Subjects v. subject fields (pp 10-11),
Subjects as targets of organised information sources (pp. 11-13)

CHAPTER 3: DEVELOPING BSO - COLLECTING, STRUCTURING AND FEEDBACK
(pp. 14-22)

CHAPTER 4: THE FIELD TEST OF BSO
(pp. 23-33)

CHAPTER 5: APPLICATION OF BSO

The primary purpose for which BSO has been compiled is to serve as an exchange or switching language for use in an information network covering all subjects and in principle extending to users anywhere in the world.

Concept representation as the basis of switching

Behind the surface idea of subject indication switching between different indexing languages lies the assumption that despite the fact that individual centres participating in a network may differ from one another in the formalisms of their local indexing languages, there is between them an underlying agreement as to the nature and relations of the concepts represented in the local indexing languages. In other words, diversity belongs to the plane of language and terminology, but agreement to the plane of thought and idea. Switching is accordingly feasible on the plane of thought and idea on which agreements exists.

Different sets of indexing terms, descriptors, or notation symbols, used in different indexing languages to represent the same idea can be made to switch their idea-content between centres, provided that

a) each local indexing language consists of terms and symbols, each of which is the sole representation, in the language, of a particular idea, and also represents that idea alone

b) some neutral representation of the idea, agreed by all concerned, becomes the medium for clerical linkage for switching purposes. The neutral or switching language of concept representation must, like the local languages involved in the switching process be a controlled language.

Does this mean that a centre using free-text indexing cannot participate in switching? The answer is that in formal terms such a centre could participate, but practically it is unlikely to do so, because, in preparing the necessary concordance tables between its own input and the switching language, it would need to embark upon a vocabulary control exercise no less onerous than the control of the local indexing language itself: this is, however, the burden from which free-text indexing seeks to escape.

 

Form of switching language

The next question which arises is: what form of controlled indexing language is appropriate for the switching duty? Should it be arbitrary identifying code, a thesaurus, or a classification? An arbitrary code which carries no implicit or explicit information upon relations between vocabulary control itself - namely, the selection of codes to represent concepts uniquely - depends upon prior process of clustering concepts in order to establish near relationships and actual identity. An arbitrary code is no aid to such clustering. ON the other hand thesauri and classifications do display semantic relations - relations between ideas on the plane of meaning.

The choice between universal classification and universal thesaurus for the switching language role follows from the manner in which each displays relationships. A classification attempts to display relationships as a totality by means of tabulation. A thesaurus depicts relationships in a fragmentary manner, in the form of binary linkages, each of which is probably separated from semantically 'next neighbour' binary linkages by the accident of the alphabet. There is very little question as to which manner of relational display is the more useful for the purpose of controlling the vocabulary in face of an incoming flow of candidate new terms. Indeed, it is becoming increasingly common for thesauri themselves to supplement the fragmentary manner of showing semantic relationships, by adding to the alphabetical sequence of keywords, ancillary sections of grouped, categorised, or fully classified terms. Indeed it is becoming increasingly common for thesauri themselves to supplement the fragmentary manner of showing semantic relationships, by adding to the alphabetical sequence of keywords, ancillary section of grouped, categorised, of fully classified terms. Fro a small thesaurus the clustering process essential to vocabulary control in admitting new terms may be undertaken informally as a purely mental activity. If the thesaurus is large and of wide subject scope, then reliable and economic control of the vocabulary requires that the clustering should be externally formalised as a classification structure. It has been argued earlier that the practicability of a universal switching language depends critically upon its ability to be controlled, revised, and updated with minimum effort. A classification , more than any other form of indexing language, is amenable to easy, predictable, yet at the same time fully controlled updating. This is the essential ground upon which it is the preferred form of indexing language for the universal switching application. That existing universal classifications have failed, or are visibly failing, precisely in this respect does not vitiate the argument. The theoretical developments in classification of the last half-century have been preferentially applied to special subject classifications. BSO is in one sense an attempt to bring many of these developments into the sphere of general classification. It seems likely that these developments, all in the direction of bringing pervasive structural patterns into general classification, may hold the key to resolving the updating/keeping-up-with-knowledge problem which besets the established systems of universal classification and their users.

New knowledge, new technology and universal classification

On the broadest perspective, the UNISIST requirement of a classification, covering all fields, for exchange or switching purposes, may be seen as a particular concrete manifestation of a more general new need for a universal classification which has emerged only in the present decade. This need has arisen from the conjunction of three separate factors. The first of these concerns the process by which growing points of new knowledge often appear astride of discipline boundaries, and in aggregate have the effect of diminishing the practical significance of these boundaries. This process has been well recognised for many years but its impact has only recently been fully felt. The fringe or marginal subjects of specialised information services are spreading ever more widely over the total field of knowledge. It is not only that some of the socially more significant of the new technologies are of mixed scientific parentage. There is at present a considerable emphasis on what may be termed holistic approaches to all departments of human affairs. The ground 'between' technology, economics and apparently more distantly related social sciences is at present receiving unprecedented attention, as may be seen from the appearance of such interdisciplinary information services as SPLINES. Equally the boundaries between technology and social sciences have become blurred by the integrated concept 'Environment' which ultimately stems from the realm of biology and psychology. The rise of this holistic standpoint has on the one side strained the capacity of the established general classifications for accommodation to near breaking point, and on the other stimulated a new need for a universal classification.

The second factor contributing to a new need for universal classifications is directly technical in character. The limitations of clerical manual methods of manipulation and transfer of information records tended to confine such activities to single disciplines, within which quantities of material to be processed were sometimes manageable. Electronic data processing has vastly relaxed these limitations, and accordingly the significance of the discipline boundaries themselves has relaxed.

The third factor contributing to a new need for universal classifications is directly technical in character. The limitations of clerical manual methods of manipulation and transfer of information records tended to confine such activities to single disciplines, within which quantities of material to be processed were sometimes manageable. Electronic data processing has vastly relaxed these, limitations, and accordingly the significance of the discipline boundaries themselves has relaxed.

The third factor leading to a renewed need for a universal classification has been the internationalisation of information processing activities. Access to information is far less than hitherto the prerogative of advanced countries alone. In the developing countries there is at present great activity in the setting up of information centres covering all fields of knowledge, and collecting or arranging access to information from all sources. This flow of information on a global scale has re-animated the whole issue of a universal classification, particularly in its role for indicating the nature of the subject-content of information requests and documents.

 

BSO for switching and mediating

All of the three above factors are clearly related to the universal switching language application of BSO. An information network should be capable of connecting centres individually oriented to different focal disciplines. Its practicability on a large scale depends substantially upon exploiting data processing and transmission technology, and the associated switching language has to be capable of surmounting linguistic and cultural barriers.

There are other possible applications, essentially of the same operational type as the switching language, which may be envisaged for BSO. In all cases they are products of the first and third of the three general factors mentioned above which seem to require a new universal classification, but the second factor concerning the liberation of earlier restraints owing to data processing technology is generally less significant than in the switching language application and may be absent altogether. In most cases, though not exclusively, these additional applications involve users who potentially may be in any part of the world.

Networking is not the only context within which neutral mediating languages come into play. Considerable financial resources are at the moment being applied to the translation, harmonisation, and interconversion of thesauri. For the minority of thesauri which themselves are no more detailed than BSO it is possible to conceive of BSO as a clerical switching language. For larger thesauri involved in interconversion projects designed to give users of one indexing language access to documents indexed in a different language, the use of BSO is a mediating, or common reference, language would achieve the necessary preliminary clustering of related terms from both thesauri, and would provide a framework for the higher organisation of the formed clusters, which might not be carried through to the finished product, but in any case would be useful as a provisional concept-holding device while the conversion work was in progress. The advantage of this approach would be both to eliminate decision process in the preliminary clustering, and to enable the broadest view of the overall subject-structure of the thesauri to be available from a very early stage in the project. Thus costly looping back whereby an early decision has to be modified to conform with the implications of a decision taken later - very characteristic of piecemeal operations on a structure of which the integrity is for the time being invisible - would be eliminated. The preliminary clusters thus formed would of course require to be broken down further by human intelligence - this being the inevitable limitation of a 'coarse' ordering system.

 

Other applications

Another example of a possible application of BSO is as an aid in the routing operations of referral centres and clearinghouses in dealing with inquiries. Compared with the switching application, this use of BSO would exploit communications technology equally, but its involvement with data processing would be less sophisticated.

Ultimately serving the same purposes as the referral centre, but serving individual demand by the mass medium of an older form of communication - the printed world - is the comprehensive directory of specialist organisations and specialist information sources. From the point of view of subject indication, present standards in publications of this kind could be improved to the substantial benefit of users. Such improvement could be realised either by arranging the material by BSO codes or by providing and index from BSO codes to page or item serial numbers.

It is also possible to envisage the use of BSO in purely disseminative modes of communication. As a subject tag supplied on copies of distributed reports and separates of all kinds, BSO codes would serve recipients of this material both as a 'coarse' interest filter, and secondly as a temporary filing system both for purposes of retrieval and subsequent control of disposal of little used material.

In these latter applications, and in some others such as the possible use of BSO codes as subject indicators in machine readable records. BSO would to some extent be competitive with existing established general classifications. The seriousness of this competition would perhaps depend upon

a) the inherent advantages and disadvantages, input cost-wise and user-wise or relatively 'coarse' subject specification versus the more detailed specification aiming at book level or documentation level in the established schemes

b) the relative merits of BSO and the established schemes in providing unequivocal placing for subjects, and thus in ease of decision effort in indexing

c) achievement by BSO of a new style of updating arrangement which would permit prompt assimilation of new knowledge into the scheme at a cost to the user which would be found acceptable

One final question which arises here is whether BSO might conceivably in future infiltrate or invade the territory proper of the established document classifications. In other words, will it ever be used for shelving books or filing documents in libraries? The answer to this question depends upon established systems rather than upon BSO. All that can be said is that if the established systems are found wanting on either of the two issues labeled b) and c) in the foregoing paragraph, then this same question will doubtless be raised repeatedly. It is not entirely unusual for tools of this kind to be used for purposes other than those for which they were originally intended. Furthermore there is nothing in the design of BSO which would inhibit elaboration to a greater depth of detail.

CHAPTER 6: BSO - DESCRIPTION OF THE SCHEME

One System design and user effort

It has been suggested in the preceding chapter that a switching indexing language needs to be economical in usage. The benefits of networking are not obtainable entirely without cost. The indexing of material by a switching language at a centre would, after all, be an addition to indexing effort normally put forth for local purposes. It is therefore essential that the additional cost of communication with other centres in the network should not contain any unnecessary element. It is against this background that the question of the cost of BSO to the user, both in day-to-day operation and in making changes consequent upon changes in the content and structure of knowledge, has been a matter of primary concern at every step in designing the scheme.

A classification user's unnecessary costs arise mainly in two ways. First, day-to-day application of the scheme may demand more decision effort than is necessary. Second, the local implementation of update amendments to the scheme may involve unnecessary effort.

Unnecessary decision effort is the result either of gross mismatch between the subjects found in the material to which the classification is to be applied and the concepts represented in the classification itself, or to lack of structural homogeneity in the scheme itself. It should be noted that mismatch is the result not only of initial shortcomings of the scheme but also of delays in updating. Lack of structural homogeneity may be paraphrased as unnecessary complexity in the scheme due to absence of overall pattern. An example of an inhomogenous general classification would be one which was prepared simply by bringing together the special classifications corresponding to each included subject area, and listing them sequentially (possibly in some logical or otherwise helpful order). Any discipline, almost by definition, represents a particular viewpoint. A series of classifications, each optimal for the needs of a particular viewpoint, form, when added together, a general classification of great complexity, and consequently demand excessive decision effort in being applied.

Unnecessary effort in implementing updating, both on the part of the updater and of the. user, is demanded when the insertion of a new subject requires not only an addition to the schedule but also a re-notation of adjacent terms representing old knowledge. This may arise either because the area involved was in the first place inadequately structured or because of a constraint offered by the notation.

These considerations are reflected in the general features of BS0, which include a marked incidence of structured pattern, both within and transcending discipline boundaries. The system is also highly prescriptive. There are no alternative placings offered. Completely definitive and embracing procedures are laid down by which indexers deal with the necessary factor of cross-classification in the schedule, which is therefore expected to be non-ambiguous in use and predictable in updating.

Neutrality and value judgments

After the question of the economics of the system comes the matter of its neutrality. All special classifications reflect the special viewpoint partly inherent in the discipline concerned and partly conventional among specialists within the discipline. Likewise, all general classifications are vulnerable to the charge that they reflect some particular world outlook or philosophy. This is obviously a question with potentially serious implications for a scheme intended for global use. Like the material which will be subject to switching in the foreseeable future, BSO reflects in many ways the standpoint of European tradition and culture. Within this limitation, the compilers have tried to stand outside sectional philosophies and to avoid decisions which have sectional philosophical implications. It is perhaps necessary to insist that neither the hierarchical nor ordinal position of any term carries any implication as to the importance of the associated concept.

Outline of BSO

The outline of the system is as follows:

FIRST OUTLINE OF BSO

It is perhaps instructive to compare this outline with those of two other systems to which it has some resemblance. The first of these is the draft arrangement of subject fields according to Object Areas, devised by Dr. Dahlberg and detailed on pages 16-17. As has been made clear in Chapter 3, this was undoubtedly a major germinal influence upon BSO. The second scheme played no special part in the development of BSO; rather, likenesses to it gradually emerged as BSO was progressively elaborated. This is the outline of the Bliss Bibliographic Classification, which has become of topical interest since the first volumes of the 2nd edition of this classification (BC2) appeared in 1977. While a draft outline of BC2 had been available during most of the BSO development period, no special significance was attributed to it during the Working Group's review of sources of terms.

All three schemes (BSO, Object Area Scheme, and BC2) are fairly similar down to point 300 in the BSO outline. Both BSO and BC2 begin with Generalia and Phenomena from a multi-disciplinary point of view, though they disagree as to which of these two should appear first. The preliminary sciences (112 to 188 in BSO) are virtually identical in content with a similar group in BC2, though the three subgroups into which these fields may be divided are arranged somewhat differently in the two schemes. Only one of these subgroups occupies a similar preliminary position (Area 1) in the Object Area Scheme. The other two subgroups are located in Area f1 to which there is nothing comparable either in BSO or in BC2.

Both BSO and BC2 have Science and technology (taken together) immediately following the above-mentioned 'preliminary' sciences; all three schemes have General technology remotely located from Natural Sciences and closely following Economics and Management of Enterprises. BC2 and the Object Area Scheme are alike in intercalating Physics-based technology and Chemistry-based technology with the respective broad sciences. The Object Area Scheme goes further in placing Mining and Metallurgy within the Area of the Cosmic and Earth Sciences. BSO, alone of the three, after assigning a place near Natural Sciences for Science and technology taken together and generally, places all Technology based on Physics and Chemistry in a position remote from Physical Sciences. This position corresponds to that of general Technology in BC2 and the Object Area Scheme.

All three schemes agree in forming a ladder of the sciences, exhibiting Henry Bliss's 'gradation of specialties' and corresponding fairly closely to an arrangement of entities illustrating integrative levels of increasing complexity. This sequence begins at Physics, passes through Chemistry, Space and Earth sciences to Biology and Medicine, Psychology and Education. BSO and the Object Area Scheme go a little further together by adding Sports, Games and Leisure after Education. In BSO alone Sports, Games and Leisure are comprehended under the more general idea of Human Needs.

In the Life Sciences area BSO and the Object Area Scheme are in approximate agreement in placing Agriculture and Animal Husbandry in the same general region as the parent sciences Botany and Zoology. BC2, on the other hand, separates Agriculture and places it under remotely located General Technology.

In the area of the Humanities and Social Sciences the divergence between the three schemes are greater than in the Natural Science area. BC2 and the Object Area Scheme both place some, but not all, of the Social Sciences before History. BSO reverses this, thereby keeping all Social Sciences together, with the exception of Education. The end of the BSO outline is generally similar to the corresponding part of the Object Area Scheme. BC2 is here sharply different from the other two schemes in placing Religion, the Occult, and Ethics before the more 'practical' Social Sciences, and also in locating Recreative Arts immediately before Fine Arts.

Taking now the broadest view of the order of the BSO outline, the 'preliminary' sciences (112 to 188) are essentially methodological sciences and techniques, applicable to many fields, and necessary tools for activity in the subject fields 200 to 890 (with the probable exception of 510 History and 520 Area Studies). It has already been noted that the sequence from 210 Physics onwards is one of increasing complexity. As this is of practical consequence in connection with matters of citation order in classifying composite subjects, it may bear restatement in different terms. Each science in this sequence has methodological and phenomenal aspects which when taken in isolation belong to preceding sciences in the sequence and not to following ones. Conversely each science in the series may contribute 'aspects' to sciences following it, but not to those preceding it.

The fact has already been mentioned that BSO follows a more traditional line than either BC2 or the Object Area Scheme in completely separating the applications of physics, chemistry, and of the space and earth sciences, from the parent sciences. It should be noted that no general classification since the Subject Classification of JD Brown has attempted a completely intimate collocation of sciences with their applications. No one now advocates this intimate collocation which fragments the whole of science and the whole of technology. The BSO Panel saw virtually no advantage and very many disadvantages in separating the steadily converging physical sciences by inserting their associated technologies between them, nor in the concomitant scatter of technology, with General Technology occurring later in the schedule than many individual technologies. However, in the area of the Biological Sciences both BSO and the Object Area Scheme collocate sciences and their applications at a very broad level only. For instance BSO has Botany and Zoology followed by Agriculture comprising Plant Crops and Animal Husbandry. The more intimate collocation giving the sequence Botany - Plant Crops - Zoology - Animal Husbandry is rejected by both schemes. When, on the other hand, we reach the Biomedical Sciences both BSO and BC2 have the science and technology closely intermixed, so that, for instance, the physiology and the clinical medicine of a particular body organ are placed together

Syntactic aspects and combinatory facilities

The discussion of subject sequence in the BSO schedules so far undertaken in this chapter has been against the background of semantic relationships - the closeness or distance between terms on the plane of meaning, when they are considered simply as isolated terms. However, so-called cross-classification has been mentioned in passing. It has also been stated that some subjects may be used as 'tools' in other subjects, and that some contribute 'aspects' to others. 'Tools' and 'aspects' represent certain kinds of syntactic relations. These are relations, at the concept level, between terms which stand together to denote compound or composite subjects. 'Cross-classification' is frequently used to refer to the dilemmas experienced by classifiers attempting to assign places for composite subjects in classification schemes which are inadequately prescriptive on the handling of syntactic relations.

Among organised information sources there are some which are devoted to subjects which are composite in nature. Accordingly BSO has comprehensive facilities for combining notational elements to represent composite subjects. It is, in fact, a fully synthesising or faceted system, though it has not been thought necessary or even desirable to label facets as such.

Combinatory facilities in classification systems inevitably raise the issue of order in which the elements are combined, also called citation order or facet order. In some working situations this issue may be bypassed or left to intuitive judgment, but for a neutral mediating indexing language covering all subject fields a completely fixed or prescriptive citation order appears to be necessary to ensure reasonably' noise-free transmission of information.

In BSO the order in which notational elements are combined to form codes for composite subjects is in the majority of cases the reverse of the order in which the elements are set down in the classification schedule. Without the qualification in the majority of cases citation order problems would be reduced to purely clerical procedures, and if we can specify those situations to which the reverse-schedule-sequence rule applies without exception, we still have a highly time and effort saving feature of BSO.

It is first of all useful to categorise combinations into internal combinations which comprise notational elements drawn from the same subject field (e.g. 575,32,0,73,50 Child welfare in disaster relief, constructed from the elements 575,32 Child welfare and 573,50 Disaster relief and aid) on the one hand, and external combinations constructed from notational elements taken from different subject fields (e.g. 550-163 Operations research in public administration, using elements 550 Public administration and 163 Operations research) on the other.

In order to make this categorisation completely explicit it is necessary to state unambiguously what is meant in this context by a subject field. Subject fields for defining internal vs. external combinations are enumerated as 'Combination areas' on page xi of the published BSO. A combination with both elements drawn from one of these 'Combination areas' is an internal combination.

Internal combinations without exception obey the reverse-schedule sequence rule for combination order. (In the above example the leading element in the combination, 575,32 is later in the schedule than the
second element 573,50).

The structural background to this combination rule, is that each subject field is elaborated according to a facet pattern, which, with very slight variations, is repeated over many fields. The following is the commonest facet pattern, given in schedule sequence which would be reversed for combination order:

1) Tools or equipment for carrying out operations
2) Operations (i.e. purposive activities by people)
3) Processes, interactions
4) Parts, subsystems of objects of action or study, or of products
5) Objects of action or study, or products, or total systems

(In the example above the first element in the combination order, namely the concept Child belongs to facet (5), the second element, the process which requires a welfare operation to be undertaken, namely the concept Disaster, belongs to facet (3). Facet (4) is inapplicable to this subject field. Facet (2) is applicable but has no role in this combination because the operation, Welfare already defines the whole 'combination area'. Facet (1) would be applicable if a particular kind of welfare agency were to be specified). Such regularity of underlying pattern covering the whole scheme is conducive to economy both in the day-to-day use of the scheme by indexers or searchers and in predictable updating.

The reverse-schedule-sequence rule cannot be used in the same clerical or mechanical manner in deciding combination order for external combinations, though more often than otherwise it would give correct and consistent results. The reason why it cannot be employed reliably for external combinations can be shown from a single example. Let us assume that reverse-schedule-sequence is being used as the basis for combination order in the case of Educational psychology. The rule will then give 460-450 (460 is Education, the hyphen or dash is the connecting symbol for external combinations, 450 is Psychology). Educational psychology, may be approximately factored as the Psychological aspects of the Education process. How then do we code Psychological education, the teaching and training in the subject Psychology? If the reverse-schedule-sequence rule were used we should arrive again at 460-450 as for Educational psychology. For any indexing system covering the whole of knowledge this would produce unacceptable noise at output. The example leads to two further considerations. The first is that external combinations should not (in the manner of the UDC colon connecting symbol) be used to indicate any relationship. This would lead not only to output noise, but also to anomalies in file sequence of classified material. The second consideration is that both in Educational psychology and Psychological education, one of the subjects (or rather the phenomena of one of the subjects) is the 'recipient' or 'target' to which the other subject contributes a set of aspects or properties. Thus psychological viewpoints are contributed
to the education process in Educational Psychology, and an education process is contributed or applied to the realm of psychology in Psychological Education. An interesting difference to be noted in passing is that the 'recipient' in Educational psychology is the primary phenomena of education, i.e. the education process, while the 'recipient' in the case Psychological education is not the primary phenomena of psychology - there is no reference here to the educating of psychological processes - but the second-order phenomena of people involved in psychology as a field of interest or profession.

The upshot of these considerations is that combination order for external combinations in BSO needs to be determined by reference to the relation between the elements which require connection. The following rule which emphasises the directionality of the 'recipient' element in the relation and the 'aspect contributing' element is believed to be unambiguously applicable to situations which can be represented by external combinations, and is recommended:

Cite first	the notation for the element denoting application area, mission, purpose, end-product or whole system: more generally the subject which 'receives' an action or effect, or is seen according to a particular viewpoint, or has a property attributed to it
Cite second	the notation for the element denoting aspect, approach, action applied, agent, or part of a stated whole: more generally the subject which 'contributes' an aspect, approach or action.

Use of the above relational formula where the 'aspect contribution' element belongs to the area 210 to 450 will normally produce combination orders which reverse the schedule order, as in the case of internal combinations throughout the schedule. This is because in this area the entities and phenomena studied by a particular science include aspects and properties which essentially belong to other sciences located earlier in the schedule sequence. For instance biological entities may have physical or chemical properties: medical, psychological and social phenomena may have biological aspects. In these cases the roles of 'aspect contributor' and 'recipient' elements cannot be reversed, as long as the 'recipient' element is the primary phenomena of the subject field concerned. If, as has been shown in the case of Psychological Education the 'recipient' element is the second order phenomena associated with the subject field an apparent reversal of roles results in a combination order which is identical with the schedule order of the elements.

Cases in which the 'aspect contributed' element belongs to the area 460 to 992, and to which the relational formula is applied, more frequently produce combination orders which break the reverse-schedule sequence rule. However, a glance at the outline suggests that the main exceptions to the reverse-schedule-sequence rule fall into a few categories. These are

1. Any social or historical aspect of any subject field 112 to 480 ('social' here is to be understood as including any aspect corresponding to subject fields 530 to 588)

2. Terminological aspect of any subject field 112 to 890.

Finally it should be noted that some departures from the reverse-schedule-sequence rule occur when the relational formula is applied to composite subjects of which both elements are drawn from the area 112 to 188.

Discipline and phenomena classes

One further feature of the BSO outline is worthy of mention. General classifications are based primarily upon subject disciplines which are methodologies and special points of view usually, but not necessarily, focussing upon a definite set of entities or phenomena. A consequence is that in conventional general classifications there is no way of classing entities or phenomena as such, merely described, or treated from many points of view. An institution dealing with, for example, Fish in all their aspects, zoological, economic, aquacultural, technical, mythological, and as quarry in a pastime, is not appropriately placed in Zoology (345,62). Furthermore there are organised information sources dealing in a multi-disciplinary manner with such topics as Food and Housing. In virtually every general classification before BSO such vital topics of everyday life have been misleadingly assigned to the discipline Sociology. Sociology is also an invariable dumping ground for multi-aspect studies, also reflected in institutional warrant, of social groups, such as Women, Racial Minorities, the Aged and the Disabled. These studies are by no means primarily sociological in viewpoint or treatment. BSO has attempted to deal with this problem by including a few phenomenon- or entity-based classes in its main outline, all containing a marked human reference, and by supplying a special location 088 at the beginning of the classification, for other phenomena or entities not included in the above-mentioned phenomena/entity-based classes. The enumerated phenomena/entity-based classes are 470 Human Needs, covering Food, Clothing and Shelter in their most extended aspects, together with Leisure, 520 Area Studies which are multidisciplinary in character, and 528 Social Groups. In connection with the residual phenomenon/entity class at 088, the problem arises as to how the multifarious phenomena and entities which might need to be assigned here should be individualised and ordered. The solution to this problem, as used in BSO, may be illustrated by the example of Fish already given. It was pointed out that Zoology offers one point of view upon Fish and that it would therefore be wrong to assign multidisciplinary (or non-disciplinary) material on Fish within the discipline of Zoology. Yet despite this mismatch in aspect or point of view Fish have a relation with the discipline Zoology which is not of the same Kind as their relation to Economics, Aquaculture, Food technology, Sport, or Mythology. The special relation with Zoology consists of the fact that the concept Fish is uniquely defined by the zoological characteristics of fish, namely their anatomical and physiological features. The concept Fish is not similarly defined - though it may well be described - in terms of the characteristics peculiar to Economics, Aquaculture, Food technology, Sport or Mythology. When treated in multidisciplinary manner any entity such as Fish may be linked - though not subordinated - to the discipline within which it is uniquely defined, and this circumstance make available a mechanism whereby an entity may be individualised and ordered at 088. The notation for the entity within the discipline which defines it is simply added to 088,. Thus the notion of Fish is uniquely defined in zoological terms. The notation for zoological aspects of Fish is 345,62. The same notation added to 088 as 088,345,62 then signifies Fish in all their aspects, zoological and other.

A somewhat different basic view, but a similar mechanism, is applied to the problem of individualising technical products. There is no question here of multiplicity of points of view. The point of view is assumed to be technical, embracing manufacture and the technique for using and maintaining the product. The problem is simply one of individualising the great number of kinds of products which emerge from technical processes. In BSO products defined by purpose or designed for a particular purpose are classed at the end of the Technology schedule at 890, and individualised by reference to the BSO code for the particular purpose, elsewhere in the scheme. It is necessary to emphasise 'elsewhere in the scheme' as the purpose of some products is simply to contribute to more complex technology. Such products (e.g. Switchgear) with a role internal to technology are normally enumerated in the BSO Technology schedules. The scheduled heading covers both their manufacture and use (Manufacture can be distinguished from use by employment of the suffix ,06,20 taken from 620 Production technology). One consequence of the policy for individualising by purpose those products with purposes external to technology is that 877,60 Cloth and fabric technology does not schedule manufacture of clothing as a product. The technology of the purpose-defined product Clothing is classed at 890,472. The 472 is taken from the root Human Needs code for Clothing.

 

Common Facets

BSO has Time and Place facets, introduced by notation -01 and -02 respectively, which are functionally similar to Time and Place divisions provided in other general classification schemes. They are applicable to every subject field except those such as 510 History, 520 Area studies, 544 to 546 Political history and Politics of individual states and groupings of states, where Place and Time are specially scheduled facets. The Place facet makes use of ISO two-character alphabetical codes, and can also specify transnational political areas (e.g. EEC countries) areas defined by language, race or religion, and areas defined by the usual physical geographical factors (e.g. Tropical areas).

An Optional facet enabling the type of information source to be specified has also been included as a result of the field test. It was found that data as to type of information source is often given prominence in descriptive material upon which indexers rely in order to establish the subject field for classifying purposes. However this data does not form part of the subject description and failure to realise this results in codes being applied which give misleading information. For example, lack of attention to this factor could cause such an information source title as British Technology Index to be wrongly coded as

600-026,GB An information source on the technology of Great Britain

whereas the correct coding is

600 33-026,GB An index, originating in Great Britain, on technology

Were it decided not to use the Optional facet, the correct coding in this case would be 600. Though the above example is taken from the area of technology, ambiguity in the use of place designation in the titles of services and institutions is even more commonly encountered in the social sciences. The use of the Optional facet compels the classifier to penetrate such ambiguity in searching for the correct subject description of an item.

Notation

Notation is the last feature of the BSO to be dealt with in this descriptive account of the scheme, and this perhaps reflects the view of the compilers that notation is at all times to be regarded as an ancillary to the structure of the classification. The scheme was in the first place constructed independently of any notation. The present notation could be uncoupled and another used in its place without changing the character of the system, always assuming that any new notation would be no less able than the present one to handle combinations and produce the required order.

The notation given in the published BSO is intended to be read and carried in mind by human users. There could be good reasons why a notation intended to be read and stored by a machine might be rather different. The human user reacts negatively to overlength and over-complexity arising from the appearance of symbols from different sets (e.g. alphabetical and numerical). Within the necessarily prescribed size limits of even variable length records, the computer is not seriously troubled by length of notation, and as all species of digits are in any case converted to numerical values for processing, a superficially mixed notation has no terrors for it. The BSO notation is believed to be tolerably brief: over 90% of the uncompounded terms cited in the schedules have codes of 5 numerical symbols length. By relying on the use of numbers as the main symbol set, and using other symbols only sparingly it manages not to he over-obtrusive. Also, by eschewing the secondary functions often accorded to notations it is capable of admitting new subjects, without limit, at their logically correct positions. It fulfills its primary function of mechanising the sequence of subjects in the schedule or in a user's file but it gives no structural information apart from that necessarily implied by the order of subjects alone. Notations of this kind are often termed 'non-expressive' notations, though it should not be overlooked that such notations do express syntactic relations. A 'non-expressive' notation was devised for 850, because all experience of the earlier established general classifications goes to show that notations which express structure, particularly hierarchy, create very difficult and sometimes insoluble problems in the insertion of new subjects in their correct places. Too often new subjects are inserted in the wrong place because of the presence of a notational gap, or the process of inserting it in the correct place involves the re-notating of the neighbouring part of the schedule. It is the dilemma embodied in these two alternatives which causes most of the decision effort and cost entailed in revising the established general classifications. This effort, and the associated delay and cost to users, should not, it was felt, be accepted as a necessity in connection with an ongoing universal switching indexing language. It was here, more perhaps than anywhere else, that the requirements of the UNISIST switching language demanded a complete break with tradition.

It was stated above that the published notation was intended for the human user. This is not quite the same as saying that it is intended only for manual switching systems. Computer processed switching systems also have human users of switching languages at both input and output ends of the switching system. Also, it is not to say that the notation could not be fed to a computer for switching between symbols having the same meaning in different local indexing languages. However, if facilities for combining switching with interactive computer-aided search were required, it would be preferable to employ for this purpose another notation containing built-in cues enabling the machine to traverse requested search paths. For the human user, the schedule of terms itself, the conceptual pattern implicit in the manner of their ordering, their hierarchical status, and the cross-references directing to related locations in the schedule, together constitute the search aid. For computer search all these matters must be explicit in the notation. Such a fully-expressive computer-oriented notation would be far too long and complex for direct use by human beings. However, given removal of the constraints upon length and complexity necessary for the human user, such a computer-oriented notation could be as hospitable to new knowledge as is the present BSO human-oriented notation.

Arabic numerals were chosen as the base symbol-set of the BSO notation because they are the best known set with elements carrying well-understood sequence values, and because they are invariable throughout the world. The numerical characters are supplemented by two punctuation signs, the hyphen and comma, and by the Roman alphabet A to Z for occasional situations where individualisation rather than grouping is required, as for instance in specifying the names of individual artists. Some notation elements are drawn from outside coding systems, such as the ISO code for names of countries, and the Groups of the Periodic table also employ Roman numerals. The use of characters supplementary to numerals demands a fixed system of ordinal values as between the supplementary characters and numerals. The following sequence gives the recommended ordinal value system for files organised by BSO:

Spaces after last symbol of notation
Two spaces, followed by further numerical characters	This occurs when the Optional facet for type of source is used.
- followed by further numerical characters	This is the connecting symbol for external combinations of notation.
, followed by further characters	This is a semantically empty character which introduces intercalated numbers filing between consecutive members of a notational array.
00 to 99 or 000 to 999	These two set never occur together in a file in such a way as to require ordinal preference between them.
A to Z

Turning now from the ordinal value of individual symbols to the make-up of a notation or code for a given subject, all codes begin with a member of the millesimal array 000 to 999. Between any two subjects represented by consecutive members of this millesimal array, further subjects may be interposed by adding to the first of the two consecutive numbers concerned a comma followed by a member of the two-digit centesimal array 00 to 99. In similar fashion further subjects may be interposed between consecutive numbers of the 00 to 99 array, by adding a comma followed by members of a further 00 to 99 array. Accordingly a typical code structure comprises a single group of 3 numbers followed by an indefinite number of groups of 2 numbers, all groups being separated by commas(e.g. 915,15,50.....)

In a few well-defined situations this typical 3,2,2.... pattern may be varied. In notations which contain the hyphen (external combinations, and Time and Place Facets) the 3,2,2.... pattern may appear on both sides of the hyphen. However, in many cases the hyphen links two groups of 3 numbers (e.g. 642-580 Nuclear reactor economics). Internal notation combinations contain the single separated number a as a connecting symbol (e.g. 978,0,72,37 The Koran). The numbers 088 and 890 are the leading number groups of untypical 3,3,2,2,... patterns.

 

Notational combination

A tabulation of the procedures for combining notation is given on page xiii of BSO, and is reproduced here on page 51. It should be added that in some fields where composite subjects are expected to arise frequently by comparison with unitary enumerated subjects, the schedules themselves provide for intersecting concepts by Expand like instructions. Notations produced by this mechanism are always shorter than the combined notation in which elements are linked by the connecting symbols mentioned in the tabulation. In deciding when to use Expand like instructions, the BSO Panel were obliged to balance the advantage of brevity against the two disadvantages that the Expand like mechanism is more likely to lead to indexing errors than the connection of two notational elements by a set of connecting symbols, and will also use up more of the available brief notation, thus in the long run causing an increase in the length of notation of future enumerated subjects. At the level of notation manipulation the difference between an internal combination (outside the area 600 to 890) and an Expand like instruction is that while the internal combination adds a connecting symbol and deletes the first numeral of the second notational element, an Expand like instruction omits any connecting symbol and at the same time deletes two or more of the first numerals of the second notational element. At the concept level Expand like instructions can be more versatile than internal combinations. This versatility is manifested in the BSO schedules where an Expand like instruction adds the legend 'with incorporated additions marked +'. These 'incorporated additions' are concepts which arise only in subordination to a facet combination. For instance the BSO Physics schedule consists essentially of a list of energy interactions and forms (from which the notion of a specific medium is absent), followed by a list of media or forms of matter. The Expand like instructions cause notions of forms of matter to be combined with notions of energy. Thus for instance we have 224,25 Plasmas and fluids, Mechanics. An important branch of Plasma and Fluid Mechanics is Magnetohydrodynamics. This branch of mechanics is logically dependent upon the combination of Plasmas and fluids with Mechanics. At the level of generalised energy interactions there is probably no term to comprehend the abstract idea of mechanical motion, magnetic fields and electric fields in triangular interaction. Accordingly 224,34 Magnetohydrodynamics is entered as an 'incorporated addition' subordinate to 224,25 Mechanics of plasmas and fluids.

REFERENCES

1 BLISS BIBLIOGRAPHIC CLASSIFICATION: 2nd ed. Edited by V. Broughton, and J. Mills. London, Butterworth, 1977- in progress

2 BLISS, H.E. Organization of knowledge in libraries and the subject approach to books. New York, H.W. Wilson, 1933

3 BROWN, J.D. Subject classification. 3rd ed. revised and enlarged by J.D. Stewart. London, Grafton, 1939.

Procedures for combining notation

For composite subjects not given in Schedule and not derivable from 'Expand like'notes in schedule

CHAPTER 7: PRACTICAL SUBJECT INDICATION WITH BSO

This chapter will deal with some general issues which arise in the practical use of BSO as an indexing language. Initially its application as a subject tagging or ordering code for information-bearing items recorded in a static medium, such as a directory, will be considered. In applications of this kind the linking of BSO codes to local indexing language codes is not a factor which has to be taken into account. When such linking is necessary, as for instance when BSO codes need to be linked to indexing languages of specialist information centres in an exchange network, fresh considerations come into play. These are discussed in the latter part of the chapter. [...]

Using BSO as a direct classification

BSO is controlled indexing and retrieval language. [...]

Background knowledge for concept analysis

How much background knowledge of the subjects concerned is required for concept analysis? [...]

Issues of policy in classifying

Decisions are needed upon certain areas of indexing or classification policy, which the classification system itself should not attempt to take, nor even to advise upon, because the answers are strictly related to the particular uses to which the scheme may be put. Answers are, however, needed if consistency and uniformity of practice are to be achieved. These qualities are at an especially high premium in any scheme of co-operative subject indication, of which a switching network is but one example. [...]

( This text is scanned from an old book with poor printing quality. Text conversion involved lots of manual editing/corrections, -- some mistakes may still be there. Please send comments and questions regarding the choice of text to aida.slavic@ucl.ac.uk)