1The different post-colonial periods in the 20^th century in South Africa saw intense and territorial engineering of a unique nature for the management and exploitation of spaces, first for segregationist purposes and then aiming at cohesion and unification. Beyond the general trends, the diversity of the strategies according to period, scale, environment and heritage can be distinguished. Exceptions and even aberrations that may be shown up by the systematic compilation of changes in status and boundary can take on meaning by way of comparison with other situations. Here, Dysturb provides information on the creation or appearance of new entities, and enables a periodisation reflecting the succession of models and political techniques applied to the subdivisions of spaces and representations of spaces in relation to their functional affectations and their hierarchisation. The periodisation that appears for the South African territorial engineering process thus highlights breaks or hiatuses formed by the the establishment of apartheid, and in particular the Group Areas Act, which was fundamental in systemising urban segregation and the differentiation among quarters and urban localities on the basis of race;

• the establishment of the 'grand apartheid' system, with the formation of the homelands from 1970. Technically, this was a “consolidation” of the native reserves inherited from the colonial period, aimed at creating districts on an ethnic basis, backed up by forced removals of population. These ethnic districts were aggregated to form the homelands,

• emergence from the apartheid system and complete territorial reorganisation, at provincial level, with the disappearance of the homelands and their integration into the provinces, and at municipal level with an exhaustive two-level “wall-to-wall” grid (with the exception of metropolitan areas). At provincial and municipal levels, the aim was to form entities of spatial and social cohesion bringing together the privileged areas (white provinces and municipalities) and the underprivileged areas (ex-bantustans and townships) inherited from the apartheid system.

2Further information

On the subject of the heritage derived from territorial engineering and the historical-political geography of South Africa, the ground-breaking work by A.J.Christopher, and in particular his Atlas of apartheid, constitutes the essential reference. A large body of study generating different publications (books and journals) has been conducted on the post-apartheid territorial reforms, in particular at provincial level, positioning the issues in the long term. The French-South African team1 (IRD-UKZN) formed within the framework of French-South African scientific cooperation and funded by the French foreign affairs ministry, linked to the development of the Dysturb database, has undertaken work on the models, the references and the central issues that relate to territorial subdivision and the selection of local and regional seats of government (see bibliographical indications at the end of this booklet).

Overall architecture and the higher levels: provinces, homelands, regions

3The outlines, the names, the functions and the numbers of territorial entities altered constantly in the course of the different post-colonial periods: the segregationist period, the apartheid period and the post-apartheid period. Nevertheless, despite the upheavals brought about by territorial engineering, the overall architecture of the main data collation frames for census was more or less stable with regard to the number of levels over the whole 20^th century (1911 to 2001): there were three main levels: places, districts and provinces. In contrast, status, boundaries, denominations and numbers of these entities evolved considerably.

Table 1 - Spatial frameworks used in the South African censuses. The entities selected for the Dysturb database in bold characters

Note **ii

4At the top of the scale, there were four provinces up to 1960 (the former colonies of the Cape and Natal and the former republics of Transvaal and Orange Free State), and these made up a homogeneous grid. In the ‘grand apartheid’ period, the former native reserves in these provinces formed the basis for the homelands (at first 9 in number and later 10, of which 4 were pseudo-independent) which appeared from the 1970 census onwards to form, with the rest of the provinces, a heterogeneous grid at the top level. From 1996, this collection of “white” provinces and ban-tustans disappeared in favour of a homogeneous subdivision into 9 new provinces. These were made up of the former provinces of Natal and Orange Free State and the breakdown of the former Cape province and Transvaal into 3 and 4 new entities respectively. The homelands were integrated as such into one or other of the new entities, preserving their outlines. Only Bophuthatswana was subdivided among three of the new provinces, and its breakdown led to a rectification of the historical border between Northern Cape province and North West province, the southern boundary of which corresponded to the former boundary of Cape province.

5At the next level, secondary distinctions defining territories, sections and regions appear in the censuses. These intermediate divisions, which are not retained in the database, were either technical statistical subdivisions with no political or administrative meaning, (the census sections of 1904, and the “economic” regions of 1960), or distinctions concerning certain territories with a particular mode of administration, mainly comprising native reserves. These non-permanent intermediate subdivisions are interesting inasmuch as they appear to anticipate the political and administrative subdivisions that came into force later. The distinction that held from 1904 to 1960 regarding a few “native” territories (Zululand, Transkei) was at once an inheritance from the treaty states (Bechuanaland, West Griqualand also appear as particular sections of the Cape colony in the 1904 census) and a first step towards the creation of the homelands on the 'grand apartheid' period, by way of aggregation of the native reserves.

6The 66 economic regions defined in the 1960 census are derived from an aggregation of the magisterial districts that does not draw away from the historical basis of the former treaty states2; however these regions did not form blocks in which reserves were in a majority position. In this sense, this technical subdivision based on regional specialisations is a precursor of the division into large development regions (1982) identified by letters (figure 4).

7These regions went beyond the boundaries of the homelands on the basis of polar attraction of urban core on tributary areas in an attempt firstly to establish a territorial development scheme making the best possible use of South African resources, and in particular the human resources of the peripheral areas constituted by the homelands, and secondly to manage services at regional level, distinct from the political order of the ‘grand apartheid’. Here can already be seen the principles of a centre-periphery association that were to become the basis of post-apartheid patterns, established no longer with a view to exploiting resources but in a perspective of spatial solidarity and redistribution. Thus the 9 development regions in fact formed the basis for the post-apartheid provincial divisions (Khosa & Muthien, 1998; Narsiah & Maharaj, 1997; Maharaj & Narsiah, 2005).

8The upper level of the provinces (and homelands) and the intermediate level of the regions, which was discontinuous, unstable and impermanent, both in fact correspond to aggregations of districts. The census districtswere based on the evolving magisterial district subdivisions. These changed in number and contour according to successive patterns and strategies that the database systematically takes into account, and that will be discussed below, but they do form a typical framework for the whole of the 20^th century. It is in these districts that the urban places are distinguished from the non-differentiated rural fabric. The distinction is based above all on the status of a given place, which is itself based on its origin and population rather than on its size. The list lengthens in the course of the 20^th century, and at the same time the principles governing their distinction evolve (see details on the subject of places in the first part).

The district: from the magisterial district to the integrative municipality

9From the 18^th century until the end of the 20^th century the political-administrative division of the South Africa space was performed on the basis of judicial districts known by different names until 19113, and then as magisterial districts. Each unit was associated with a capital town centralising the jurisdiction. These magisterial districts held as the basic local subdivisions for colonial and post-colonial organisation up to the establishment of the new system of local government instating municipalities in 2000. The 20^th century population censuses and the organisation of local administration used this grid, although there were variations over time. Thus the census districts consistently corresponded to the MD, but in some cases came into use with a time lapse in relation to their administtative creation. This is the case with certain MDs in Cape province created at the end of the 19^th century (Simonstown in 1848; Steytlerville in 1899; Kuruman in 1896 with the districting of British Bechuanaland; Ignawuma annexed in 1896 in north Zululand) which do not appear in the first census in 1904, but only in the 1911 census. Later there was an anticipation of the magisterial districts by promotion of several new jurisdictions and related capitals via decisions already reached in the census year and often coming into force on publication (in the 1936 census the district of Keiskammahoek was already present a year earlier, the district of Ndwedwe appeared in 1921 for an official creation as a MD in 1937). For 1951 the apparent anticipation that concerns the districts of Delmas, Kirkwood, Theunissen, Sibasa, Oberholzer, Kempton Park, Sasolburg, Wesselsbron, Vredenburg, Welkom, and Venderbiljpark is linked to the fact that in the database it is the division of 1960 that has been used to compile the data for 1951 (these are in fact retrospective data for 1951 as they appear in the 1960 census using the 1960 divisions). Anticipation in the ‘grand apartheid’ period districts relates to the same phenomenon: thus numerous districts created within the 1972 movement already appear as census districts in 1970, alongside ethnic subdivisions of MDs into “Zulu,” “Xhosa” “Swazi” sub-districts. These sub-districts were later re-aggregated to form the Zululand, Ciskei and Kangwane MDs. The table concerning the districts (in the CD-Rom annexes) provides the following for all the magisterial and census districts: date of creation as MDs, date of appearance as CDs, denomination, successive capitals. The table also indicates the previous divisions from which they were derived, and any later changes in contour. Finally, the table enables the MDs to be grouped according to genesis typology into four categories represented in diachronic maps.

10For this typology, four main phases of districting and territorial engineering can be distinguished, with a degree of overlap. They correspond to the colonial strategies for territorial exploitation and (re)distribution of population, and form the territorial heritage of apartheid. The different modes of application in the course of these districting phases (table 3, figure 5) are as follows:

• conquest-generated divisions (1a) or the formation of the initial grid of districts that accompanied the progression of the colonised space;

• first colonial subdivisions (1b) of the initial entities;

• the progressive tightening of the grid and the promotion of new centres (2);

• the creation of ethnic districts to form the homelands (the 'grand apartheid' strategies) (3).

Table 2 - Number of districts according to type of genesis and census

11In 1911, of the 206 MDs, about two thirds correspond to the initial entities in the first divisions (1a and 1b) that accompanied the progression of colonised and annexed space. These first-generation MDs had already been reduced in area, since in 1911 about a third were already second generation MDs (2) proceeding from a tightening of the grid by subdivision, along with some reorganisation.

12All of the new creations from 1921 to 1960 (some hundred in number) followed this same logic of a tightening of the grid and promotion of new localities as seats of magistracy. From 1970, this trend slowed (around ten creations between 1970 and 1991) while a new wave of creations of ethnic MDs occurred within the ‘grand apartheid’ logic, and the formation of the bantustans (around 100 up to 1991, and about 40 temporary creations in 1970 before reorganisation). Thus in 1991 more than a third of the districts related to a colonial origin, a little over half corresponded to subdivisions to tighten the grid, forming districts on the basis of attraction poles, and a quarter concerned ethnic districts created within the 'grand apartheid' policy.

The constitution of an initial district grid organisation in the colonial period

13The mode of formation of this grid differed according to the nature of the colonised spaces. However an initial model can be seen: a new Cape colony district was formed overlapping the preceding district (then no longer on the colonial conquest frontline) and extending into newly integrated territory. Thus the preceding district was progressively reduced in area around its seat. In this way a town could shift from a pioneer or military frontline position to a rear base status, and then become the capital of a small colonial district no longer in contact with the frontier. The best examples of this are Stellenbosch, Graaff Reinet and Somerset West.

14There were numerous variants to this model from the first half of the 19^th century, at a time when the pioneer front was in contact with previous dense and well-organised occupations, that gained recognition (as protectorates or treaty states), or else were fought and annexed, or again reduced to the status of native spaces, after the model of the reserves developed in the colony of Natal (under the name of “locations”). Administrative centres could then be created ex nihilo to accompany a mode of colonisation based on management rather than settlement. In this case a small white "isolate" appears, retaining its status as a district capital, sometimes becoming considered progressively as part of the native territories as in the case of the towns in the Transkeian territories, or sometimes excised from the territory initial district when it became part of a bantustan if the town was on the district fringes. Examples are King William's Town or Matatiele, around which a settlement colony established. Alongside this, the Boer conquest went forward in the interior, engendering a race towards territorial dominion via the creation of large not very organised blocks on one side (the Republics) and a boundary formed on the other side to the south by large British protectorates, the extension of Natal and the annexation of Zululand.

15It is a first subdivision of these republics and treaty states once they were annexed that forms the second mode of colonial subdivision into magisterial districts. From the outset they were designed to divide up these territories by way of a grid, producing entities of fairly equal dimensions. But the model, defining a new district-based jurisdiction around an embryonic city as the seat of magistracy and the local capital, lacked hegemony, so that the configurations of the colonial districts, their cartography and their denominations show varying practices and different models. Five regional instances can be mentioned:

• Natal, with the first division, Natalia, (centred on Pietermaritzburg) was subdivided into large entities delineated by the main rivers, and these entities were in some cases centred (Impafana on Weenen; Pietermaritzburg on Pietermartitzburg; and d'Urban on Port Natal), in some non-centred, on the fringes (Umvoti, Umzynnyati and Upper Tugela), or on the coast (Lower Umzikulu, Lower Tugela, Alexandra, Umlalasi);

• Zululand, following the difficult period of subjugation to colonisation, was first of all broken down into competing chiefdoms, and then, once annexed, was divided into districts that returned to some of the tribal divisions. North-west Zululand was integrated into the South African Republic, and later returned to Natal, forming polarised districts: Vryheid, Utrecht, Paulpietersburg;

• Transvaal underwent a first subdivision with territories named after the local seat of magistracy, and others named according to the natural regions and resources: Waterberg, Witwatersrand, Soutpansberg. The layout of the districts is a roughly circular configuration surrounding the main mining, industrial and settlement centre, Witwatersrand. Beyond, the districts were distributed in sectors with their capitals decentred and close to the provincial core, forming a first and second ring. Krugers-dorp, Pretoria, Middelburg or Heidelburg were the centripetal local capitals in the first ring, Rustenburg, Nylstroom (for Waterberg), Pietetsburg (for Soutpansberg) and Lyndenburg were in the second outer ring in the Transvaal;

• the Orange Free State was divided from the outset (including the southern part made up of concessions and treaty states) following a regular grid organised on a network of small towns;

• the Transkeian territories were subdivided before 1904 into districts which were derived from the initial divisions into treaty states progressively annexed between 1879 and 1894: Fingoland, Tembuland, Pondoland, and Griqualand East; they however still appear in the 1911 and 1921 censuses as intermediate subdivisions comprising districts with a particular administrative status.

16The hybrid denomination of the districts illustrates the plural nature of their configurations. The names of the seats of magistracy dominate (rarely based on African place names, except in the Transkeian territories) and they were exclusively used in the Orange Free State. However other denominations occur in two forms, that relating to British culture and history on the one hand (in particular in the ex-Cape colony), and on the other hand that relating to environmental features (in particular in Transvaal and ex-Zululand (Giraut et al., 2008).

A progressive tightening of the grid

17The subdivision of the initial grid began as early as the first third of the 19^th century, but continued throughout the 20^th century (more than 150 creations between 1911 and 1996, most occurring before 1970). The driving force behind this phenomenon was the densification of the administrative cover, relating first to the promotion of new localities (rural centres), and later to that of mining compounds and metropolitan peripheries which acquired municipal status at much the same time. The mode of creation of new districts sometimes involved mere subdivision of the previous grid, sometimes the creation of a new contour by way of amputation of several existing districts (the most frequent instance). Thus the evolution of the territorial grid both reflects and accompanies the evolution of population distribution, and the South African industrial infrastructure which was considerably reshaped by the development of mining activity and industrialisation. All the districts created in this logic carry names that relate to their capital (seat of the magistracy). The case of the successive creations in Orange Free State illustrates the variety of different modes of creation, with the promotion of small rural towns between the two world wars, and from 1950 a reorganisation of the new mining basin and its surrounding areas.

18Where promotions concern secondary crossroad points, the new district encroaches on several surrounding districts (for instance Dewetsdorp and Reiz, created in the 1920s, each overlapping four surrounding districts). When the secondary centre promoted is not far from the previous district capital, it is the surrounding area or hinterland that is subdivided (creation of Vredendal in Vanrhynsdorp, for instance). Finally, there is the case of the breaking-up of a district to form several entities, such as Rustenburg in 1963 becoming 4 entities with the creation of Koster to the south, Schwartruggens to the west and Thabazimbi to the north.

19Divisions within a dense fabric of small towns (the Cape Town region) give small districts (for instance Wellington from Paarl or Vredeburg and Hopetown from Malmesbury), while a little further away in areas of less dense fabric subdivisions yield large districts such as Williston from Fraserburg. If the magisterial districts of the conquest period were practically all progressively subdivided, in contrast the division into magisterial districts corresponding to a first subdivision of the treaty states after annexation appears relatively stable, and hence not much affected by this phase of creation of new districts. This is particularly true for the ex-Transkeian territories, and also to a degree for ex-Zululand.

The creation of ethnic districts

20The creation of ethnic rather than functional districts from 1970 belongs to the 'grand apartheid' policies aimed at forming independent homelands, accompanied by the mass expulsion of black urban populations towards these bantustans according to their affiliation to the various linguistic groups.

21The first stage was the formation of territorial blocks to make up the future homelands. This process started with the 1970 census in the form of ethnic sub-districts for the Zulus, the (Ciskei or Transkei) Xhosas, the Swazis, the Tswanas, the Lebowas, the Vendas, the Basothos and the Matshanganas. These new Bantu districts were subdivisions of existing magisterial districts, and were subsequently recomposed to form a new set of magisterial districts, this time cross-sectional as for the KwaZulu district. In addition, adjustments and even major changes (as for Ciskei) were made throughout the “consolidations.”

22In most cases, the creation of districts forming the homelands was performed by aggregation of the former native reserves, which were removed from the jurisdiction of the districts in which they were situated. This process concerned the bantustans other than Transkei, which was already divided into a system of districts and which was converted as such into a homeland, with merely a few adjustments on its margins.4 Instances of major reconfiguration of the district map were seen during this phase (also affecting districts bordering on those that were directly concerned by the creation of the new ethnic districts) (figure 5).

23A few examples can be cited:

• in the north, the district of Soutpansberg was first of all subdivided into five entities which then occupied only the north-west part of the district existing at the start of the century. The creation of Venda and Gazankulu led to the dismantling of the district of Sibasa in the northeast, and the white sector of this former MD was then attached, without spatial continuity, to Soutpansberg which thus integrated a piece of Kruger park situated on the other side of Venda which was independent;

• the district of Letaba in the 1970s first of all lost several sections in the centre-north to create Lebowa, and in 1982 all the centre and east of Gazankulu was separated from it for the promotion of the mining town of Phalaborwa, one of the entrances to Kruger park. Letaba survived in residual form (with no eponymous place, and cut off from Kruger) between two fragments of Lebowa;

• Maluti (ex-Matatiele) was the Xhosa part of the original Matatiele district, while the white part was attached to Mount Currie. Thus a paradoxical situation occurred in which there was a Matatiele district in Transkei that did not include the town of Matatiele which was situated in the district of Mount Currie;

• the disappearance of the district of Mafeking: this district first of all lost its south and central zone in 1971, with the creation of Bophuthatswana. It was then in two distinct parts; the northern part with the town of Mafeking was transferred to Bophuthatswana in 1980 and became part of the district of Molopo. The southern part, with no district capital, was then transferred to Vryburg.

24The administrative reorganisation of the metropolitan peripheries was also affected by this racist territorial engineering. Large townships (Umlazi in Durban) and/or informal peripheral areas (Mdantsane in East London, Zwelisha in King William's Town, Odi-Moretele in Pretoria, or Mpuma-langa between Durban and Pietermaritzburg) graduated to the status of districts before being used to create a homeland or being joined to one. Thus the homelands, structurally, are spaces with a rural centre, sometimes dense, with their main urban agglomerations on the fringes. The ‘grand apartheid’ territorial reorganisation reinforced the dual nature of South African spaces, already organised via the municipal system and land ownership patterns. This dual nature pervaded the political and administrative spheres and underpinned all public territorial policy, whether administrative, judicial, or technical. The territorial patterns were, from then on, defined sometimes on functional criteria, sometimes on racial criteria, yet the territory was mapped out on modern, rational principles of grid and nesting. The result was that the territorial patterns in South Africa yielded a mosaic of discontinuous, profoundly heterogeneous spaces. Thus the ‘grand apartheid’ territorial policies appear in some ways as a post-modern territorial management freak.

South African post-apartheid territorial engineering

Establishment of a coherent system

25After reviewing the subdivision into provinces, and removing the former bantustans, New South Africa set out to establish a rational, equalitarian municipal system. A spatial-social mix, alongside a “wall-to-wall” type grid with nested subdivisions (province, district and local municipalities) were the main underpinning principles of this system. There are however notable exceptions that do not obey the rationale. These exceptions concern first of all the metropolitan areas which became autonomous within the municipal and provincial system, and secondly the areas on the fringes of the new provincial divisions, for which hybrid status was coined in the form of the District Management Areas (DMA) and the Cross Boundaries Municipalities (Local and District Councils) (CBLC and CBDC).

26For this purpose, a law derived from a White Paper created a mandate for an independent body, the South African Municipal Demarcation Board. This, in compliance with the 1996 Constitution, was to provide South Africa with an equalitarian local government system using an overall grid system for the territory, which had not been fully achieved in the interim system, the Transitional Local Councils (TLC) established for the first local elections in 1996, and found in the 1996 census. The emphasis, in particular in the demarcation objectives, was on the ability of the different entities to ensure basic services to their populations, which implied resorting to socially composite entities of a size that would provide the local government body with a sufficient financial base. A fundamental aspect of the policy can be seen in the wording: a need for coherent, inclusive and non-fragmented entities, opening up scope for the establishment of metropolitan areas. However a set of general criteria were also enumerated to be taken into account: environmental, topographical and physical chatacteristics, the traditional rural community areas, commuting patterns and use of infrastructures, and other administrative borders.

The fringe areas, at the heart of the South African territorial reorganisation

27One of the challenges facing social, political and territorial reorganisation in post-apartheid South Africa is the way in which it was to process its internal fringe areas, those fringes where “second rate” citizens were relegated, areas distant from the economic and political centres. The task was to rehabilitate and reintegrate them, and for this purpose two requirements, to some degree contradictory, had to be met: differences in status needed to be abolished, and solidarity, hence a social mix, needed to be introduced into the new entities and institutions. The issue of how to manage the fringe areas inherited from the past thus became crucial. The issue, after dominating the debate on the subdivision into provinces, was equally central for municipal subdivision. This focus of debate on internal fringe areas, rather than on potential centres, first of all led to the inclusion in the interim constitution of a nominative list of 1994 potential border conflicts arising from the subdivision into nine provinces. Such conflicts were to be solved by referendum and/or by the new system of local government. Some of these border conflicts, in particular that of Bushbuckridge, have been studied in detail (Ramutsindela, 1998, 2001; Narsiah & Maharaj, 1999).

28Between 1996 and 2000, the debate on the subject of the transitional local councils also focused on the way in which internal fringe areas should be dealt with. The TLCs, created in haste, were an extension of the municipal status that had hitherto been restricted to white quarters of towns and cities. While integrating non-white townships in the name of spatial-social solidarity, the TLCs were often disputed because they ignored the informal or semi-rural peripheries of towns and cities, as well as urbanised areas that had shifted beyond the boundaries of the former bantustans (McCarthy & Bernstein, 1998).

29Finally, with the “wall-to-wall” grid across the whole South African territory dividing it into municipalities, metropolitan areas and districts at the end of the 1990s, the question of the customary authorities returned, and more generally the issue of autonomy for the rural peripheries. Against a backdrop of African renaissance claims launched by President Thabo Mbeki, with wide scope for interpretation, the calls of the customary authorities, threatened and marginalised by the new local government proposals, met with popular support. Setting aside the issues of redistribution of wealth among composite municipal entities, these demands, which were for a while met by the TLCs, failed in certain cases. In KwaZulu-Natal they led to the establishment of municipal entities that were completely deprived of resources, but homogeneous in socio-economic terms. In fact this amounted to granting administrative autonomy to a certain number of displaced urban fragments in the internal fringe areas of the ex-bantustans, with their customary rural environments and in some instances some modern agriculture.

30Thus the sector of Loskop5 constituted a municipality (Imbabazane Local Municipality, KZ336) that was independent from the industrial towns of Estcourt (Umtshezi Local Municipality, KZ234) and Ladysmith (Emnambithi-Ladysmith Local Municipality, KZ232), although the mayor resided there in 2002. In Eastern Cape province, it is also the case for the municipality of Peddie (Ngqushwa Local Municipality, EC 126) which, with the sector of Glenmore, gained autonomy with regard to Grahamstown (Makana Local Municipality, EC 104) and King Williams Town/East London (Buffalo City Local Municipality, EC125).

31What we have here, in contradiction with regard to the general policies of the new divisions, is an officialisation of the administrative divides inherited from the ‘grand apartheid’ between small towns and dependent displaced urban zones. This yields the series of diagrams given in part III devoted to towns and cities, which shows the municipal phases for these displaced urbanised sectors.

32The debate on the challenges of a radical reform of the state's territorial organisation thus focused on how to deal with internal fringe areas. It is indeed true that in South Africa, whatever the scale, contrasts are extreme in the fringe areas of the entities inherited from previous periods, and that the phenomena of discontinuity were exacerbated by the ‘grand apartheid’ policies.

"Empty fringes" and "full fringes"

33On national and regional scale, it is however possible to distinguish two very different, contrasted, marginal spaces or fringes. There are indeed on the one hand what could be called “empty fringes,” often also positioned on international borders, and characterised by very low population densities, with systems of management dominated by environmental protection of the national park or reserve type, former buffer zones between an isolated South Africa and the ‘frontline states’. On the other there are the “full fringes” urban fragments that were shifted outside the contours of the former bantustans, which form a particular South African feature. They are directly inherited from the displaced urbanisation policies enacted within the ‘grand apartheid’, and the establishment of the bantustans with their peripheral ‘dumping grounds’: this corresponds to a type of urbanisation directly projected several tens of kilometres away from a central town in the form of urban fragments with no central element.

34Besides the centres of metropolitan areas and dense rural areas (Zulu, Venda and Xhosa in particular), the population distribution map also highlights these ‘dumping grounds’ (Allanic, 2003; McCarthy & Bernstein, 1998). These spaces, located on the borders of the former bantustans, also found themselves on the fringes of the administrative divisions of the new province map. This is particularly true on the borders between the provinces of Limpopo and North-West and the provinces of Gauteng and Mpumalanga. There are three patterns:

• peri-metropolitan settlement: the northern periphery of Pretoria (city of Tshwane) where the last perimeter integrates the main dumping grounds of Odi Moretele (Gervais-Lambony & Guillaume, 1999);

• settlement on the peripheries of small towns: Groblersdal and Marble Hill which integrate the displaced urban areas derived from ex-Lebowa (Ramutsindela, 2001);

• residential basins on the national borders; Bushbuckridge on the edges of Kruger park and to the north of Nelspruit (Narsiah & Maharaj, 1999; Ramutsindela, 2001).

35In other instances, these former bantustan “full” fringes are integrated into a composite province, and are today poor urban isolates on the fringes of the metropolises (Mpumalanga to the north of Durban, Botshabelo and Thaba N'Chu to the east of Bloemfontein, Mdantsane and Zwelisha to the north-west of East London, Esikaweni some way from the Empangeni-Richard's Bay axis). They are thus on the fringes within new municipalities, either metropolitan (type A municipality) such as Durban, or non-metropolitan (type B municipalities), like East London/ King William's Town, Bloemfontein or Empangeni/Richard's Bay (Giraut & Maharaj, 2003). To comply at local level with the principle whereby well-provided centres are associated with underprivileged peripheries, the new municipalities in some instances comprise entities that are discontinuous in spatial terms but functionally interdependent.

36There are composite situations in which “empty” and “full” fringes coexist. This is particularly true to the north and west of KwaZulu Natal with the complexes of parks and reserves (Maputaland, Drakensberg) which are known internationally (Unesco World Heritage) and lie alongside dense rural areas (Zululand) or areas assimilated to the dumping grounds (Drakensberg piedmont). The South African provincial borders, which perpetuate the discontinuities of apartheid at regional level, are remarkable with respect to population distribution. They present an irregular succession of human concentrations and human vacuums, which are reiterated in the grid of the new municipal entities and in their statuses.