One Place, Five Authorities

Constantine refounded the city on 11 May 330 AD. That date exists because the refoundation was an imperial act performed in front of contemporary observers who wrote it down.

The city’s traditional founding, around 657 BC, rests on later Greek tradition. Both dates are in the record and neither is wrong.

They are answers to different questions, obtained by different means, and a table with one column headed founded will hold exactly one of them without saying which.

After that the site keeps its position and changes almost everything else. Roman, then Byzantine, then Latin Crusader rule from 1204 to 1261, then Ottoman, then the Republic of Turkey after the capital moved to Ankara in 1923.

One place. Five authorities.

Three names, and not one of them is an identifier for the others.

What a place record keeps in separate columns #

The fields are not difficult to write down. They just do not fall out of the label, which is why the label gets used instead. A founding date needs its method attached. Constantine’s has a ceremony and witnesses. Byzantium’s is tradition recorded centuries after the fact. Angkor’s declaration of universal kingship in 802 AD survives in Khmer inscriptions. Those are three evidentiary bases wearing one column heading, and only the third would be described the same way by an epigrapher and a database.

Extent is worse, because it produces a number and numbers travel further than the caveats behind them. Angkor’s UNESCO-inscribed zone covers just over 400 square kilometres. That figure comes from a heritage nomination survey conducted in the twentieth century. It is not, and was never meant to be, a measurement of the ancient Khmer state.

The Silk Road shows the same split in a way nobody can miss. The inscribed Chang’an–Tianshan corridor runs roughly five thousand kilometres. The full historical overland network is usually given as something over six thousand.

The first is a modern management subset. The second is a historical object. Neither is a correction of the other, and averaging them would produce a number describing nothing at all.

A map is evidence, not a picture #

Old maps get dated the way any other artifact does, and the construction is part of the proof rather than a detail about it.

The surviving fragment of Piri Reis’s 1513 map carries a legend saying its maker drew on some twenty prior charts and eight geographic treatises, including a Columbus map of the West Indies that has not survived. Its coastline data, its projection choices and its place-name conventions are checked against the sources and techniques a compiler working in 1513 would have had.

The signature is not the whole proof.

What the map was made out of is.

This is the discipline a modern geographic field lacks entirely. A map is not a neutral picture of a place. It is evidence of a particular boundary, drawn by a particular authority, for a particular date, and every one of those three facts can be missing from the file while the shape renders perfectly.

Some places cannot supply the field at all #

There is a limit here, and it is worth stating rather than working around.

A historical place may simply have no boundary to recover. The Hanging Gardens have three live positions: a literary invention, a real Babylonian structure since destroyed, or Stephanie Dalley’s relocation of them to Nineveh under Sennacherib. No Babylonian text mentions the gardens. No excavation has found them at Babylon. There is no area figure for either candidate site, and there will not be one.

The Silk Road has no founding power and no controlling-power count, because it crossed the territories of dozens of contemporaneous states and belonged to none of them.

A blank field in those records is not a lapse in research.

It is a description of the object.

Where the transfer runs the other way #

This is the point at which the historical case stops being the harder one.

A reporting system is not stuck. It can choose a polygon, a customer-assignment rule or an inclusion list before the period closes rather than after. It can store effective-from and effective-to dates alongside the choice. The choice may be entirely contestable, and it is still inspectable in a way that a vanished structure’s footprint can never be.

So the working record is a versioned geographic key: label, polygon or assignment rule, controlling entity where that matters, effective-from, effective-to, and source. The current region name is kept as an alias, because people will search for it. It cannot do the work of those fields.

A comparison across periods then runs on the intersection of the two footprints, and the remainder is reported separately as a boundary transition with the date and the rule that produced it. A real expansion has an acquisition, an annexation or a coverage event to point at. A collision has only the comfortable continuity of the name.

Byzantium, Constantinople and Istanbul are three names for one site, and the site is the only part that stayed where it was.

A region code is one name for whatever the current inclusion rule happens to cover, and the rule is not printed on the code.

The name is what makes a series look continuous. The effective-from date is what makes it true, and it is the field nobody asks for until the comparison has already been run.