Part of Spreeuw Fontein & Toornits Kraal (Spreeufontein No. 26)

Lot B (Spreeufontein No. 26 Ptn 4)

4/26 on the cadastre · on record 1909 · show on the map

The farm's own history, the grant, the original survey, sits on that page.

Farm number: No. 26
Portion: 4
On the cadastre: 4/26, portionregistered as SPREEUW FONTEIN
District: Prince Albert
Extent, as the diagram states it: 6098 morgen 243 sq roods
Modern surveyed area: 5237.3 ha

The historic extent and the modern surveyed area are two separate measurements and are not presented here as confirming each other. The first is what a surveyor wrote on a diagram, usually in morgen. The second is what the modern cadastre draws.

Why this farm is here

A piece of a farm already in this model. 54 km from Klaarstroom.

parcel Cut out of Spreeuw Fontein & Toornits Kraal (Spreeufontein No. 26)
parcel Cut out of Vredendal (Spreeufontein No. 26 Ptn 1)

What the record holds, in order

1909 boundary change subdivision: from Vredendal (Spreeufontein No. 26 Ptn 1)
undated boundary change subdivision: from Spreeuw Fontein & Toornits Kraal (Spreeufontein No. 26)

Also bordered by, as the modern cadastre draws it

Ground that touches this farm on the Surveyor-General's current polygons and that none of the diagrams held here names. Derived from the geometry, not read off a document.

Cadastre

Surveyor-General diagram index for this parcel

Surveyor-General diagram index for the whole farm — where the original survey of this ground sits.

Research notes

cut from Portion 1 in 1909; deed 1909-83-5219

*** [4 Aug 2026, LEAD-0084 RESOLVED] THE MATCHER WAS NEVER AT FAULT. *** The polygon for this parcel carries WSTATUS='H' - SUPERSEDED. match_cadastre.py indexes superseded features by their LPI ONLY and keeps them out of the name and farm-number indexes, so nothing automatic can land on one. That is deliberate and documented in the loader: a superseded polygon is reachable by an explicit PIN and by nothing else, because a pin is somebody saying "I know, use this one". For a history site a superseded shape is often the only correct geometry there is - Modderdrift Portion 4 is the precedent. PINNED HERE, on the same evidence Modderdrift was: the extents agree - 5 223,1 ha teen 5 240,5 ha, 0,3%. The pin survives future runs of the matcher by design.

[13 Aug 2026] Renamed to the one form this model uses for every portion - the portion's own name, then the parent farm, its number and the portion in brackets - so that a name is legible in a list, a search result and a map popup without the reader having to know which farm number is which. It read "Spreeufontein Portion 4 (Lot B)" before. Nothing about the ground changed.

The research behind this page, kept separate from it: dated working notes, open questions, and what is not yet settled.

Open questions

141 people are on this site linked to nothing priority 1 · LEAD-0074
Which parcels and people exist on the site as dead ends - reachable by a list, but linked to no document, story, owner or neighbour?

What would settle it For the 106 tree-derived rows, a parentage pass merged from the family-tree on tree id - which 6 already names as the direction of travel and which would connect them wholesale rather than one at a time. For the three floating names, find the document each came from and link it.
Four portions have a polygon in geo/ that match_cadastre.py still will not pick up priority 2 · LEAD-0084
Why does the matcher miss PAR-000023, PAR-000058, PAR-000069 and PAR-000144 when their exact PRCL_KEYs are present?

What would settle it Printing the matcher's computed key beside the cadastre key for one of these four.

What the record is missing

Nobody is on record as ever having held this third order · B8 never-held
Lot B (Spreeufontein No. 26 Ptn 4) is in the model and drawn on the map, and carries no ownership record of any kind, not a grant, not a transfer, not an occupation.

What would settle it The Surveyor-General's diagram for the farm number names the grantee and the date; the deed reference on it opens the transfer chain.

Found by breaks.py reading the model against itself, not written by hand, each one is a shape of missing record, not a fault in the data. The whole set is on the questions page.