The day before, I had published an article about a hook for a dog.
The next day I walked past the library and noticed a public drinking fountain.
That was enough to make the two things collide in my head.
The city had moved clean water through pipes, treatment, testing and public infrastructure all the way to a small visible point in the street. A person could walk up and drink. A bottle could be filled.
I looked down.
A dog standing beside that person had no obvious way to use the same endpoint.
My first thought was almost embarrassingly simple:
Did we forget to give the dog water?
My second thought was better:
Bring a bowl.
And that may be the right answer in most places.
An owner can carry one. The owner can bring water, buy water, change the route, go home, or decide the dog does not need water yet. A dog beside a fountain does not create a municipal obligation. It does not prove the fountain is badly designed.
But the bowl does not make the observation disappear.
It tells us that one user reaches the infrastructure directly and another reaches it only through a human intermediary and an extra object.
That is where the question became interesting to me.
Not:
Should cities install dog bowls everywhere?
But:
When water becomes public infrastructure, who or what was the final interface designed to serve?
I had already been circling this question without quite seeing it.
In Bread, Water, and the City, I wrote about Paris because Paris makes water unusually visible. A fountain is only the small public tip of a much larger system: pipes, treatment, testing, maintenance, financing, governance and repair.
The water was already there.
The fountain made it reachable.
The dog made me notice that reachable is not a property of the water alone.
It is a relationship between water and a body.
The body changes the problem
For a standing adult, a drinking fountain may be a perfectly good endpoint.
For a dog, the last gap may be twenty centimetres and one missing vessel.
For a bat, the problem is not height at all.
A 2025 urban field study by Peyton Harper and Victoria Bennett looked at bat use of water sources in Fort Worth, Texas. Bats drink on the wing. The study found that longer continuous stretches of unobstructed water were associated with more bat activity and more observed drinking events.
That is a much stronger example than my dog.
Water can be present and still be difficult to use because the body cannot perform the action the interface requires.
The dog can usually borrow a human solution.
The bat cannot carry a bowl.
For a tree, the fountain metaphor breaks completely. Usable water depends on what reaches and remains in the root zone.
A dog, a bat and a tree are not one biological category. Treating them as one would be nonsense.
What they share is only an infrastructure question:
Water can be present without being usable in the form a living body or function requires.
Or more simply:
Supply is not access. The body defines the interface.
That is not a new scientific theory. Engineers, ecologists and landscape architects already work with these distinctions in richer forms.
What I am testing is whether putting the interfaces next to each other exposes handoffs that disappear when each discipline looks only at its own part of the system.
Paris did not simply forget nature
There was an easy version of this article that I now think would have been wrong.
Paris has a large public drinking-water system. It would have been tempting to say that the city designed water for humans and forgot everything below the tap.
The evidence does not support that story.
Paris’s Biodiversity Plan 2025–2030 calls for more ponds, planted basins, wet habitats and more natural river edges, explicitly as habitat for biodiversity.
So Paris has already looked beyond the human drinking fountain.
That correction makes the question better.
Paris has infrastructure that makes potable water directly reachable to people.
It also has policies and physical systems that make water part of habitat.
The interesting place is where those categories meet.
I tried to make that concrete at Jardin Tino Rossi.
The City of Paris page for the park explicitly lists an
Espace canin and Points d'eau potable. It also
refers to a fontaine à boire by locating a defibrillator
next to it.
So this is not an abstract comparison between a dog park somewhere and a fountain somewhere else. At one named public site, intentional dog use and public drinking water are both part of the documented space.
Then I downloaded Paris’s drinking-fountain data.
The GeoJSON copy I downloaded that day contained 1,325 water-point records, including 27 in the 5th arrondissement.
Using a deliberately broad corridor around the long park, I marked
three 5th-arrondissement assets as candidates for further checking. One
is a Mât source near the Austerlitz end. Another is a
Particulier fountain on Quai Saint-Bernard. A third is a
Fontaine Arceau toward the north-western end.
But proximity is not identity.
The park is long. The dog area, the drinking points and the address labels do not line up neatly enough for me to say which asset is the relevant dog-area fountain without a proper spatial or field check.
A dog-oriented secondary guide says water fountains are available for owners at the site. I could not find a reliable source or photograph that established a low dog basin, a dog-specific spout, or an official rule telling owners to fill their own bowl.
That is not proof that no dog interface exists.
It is the point where the public material I could recover stops answering the last-metre question.
The current defensible statement is smaller:
Paris documents both dog use and potable water at Tino Rossi, but the public material I could recover does not establish whether the dog is a direct user of the water interface or whether the owner is expected to mediate.
If somebody shows me a dog basin there tomorrow, that does not destroy the investigation.
It changes the result.
Then the question becomes: Paris already solves this interface at least at some sites — so where, why, and under what operating model?
That is what a falsifiable investigation is supposed to allow.
The interface is not hypothetical
There is also a useful counterexample.
New York City’s Department of Design and Construction describes the Washington Square Park renovation as including water fountains for humans and dogs alike.
The project history matters more than the novelty of the hardware.
A direct municipal dog-water interface is not merely a product-catalogue idea. A city has documented one as part of a public park project.
That also changes what the investigation should ask.
Once a dog-water endpoint is already part of the design, the answer is not automatically build a dog fountain.
The next questions are whether the source fits the use, how the basin drains, who cleans it, what happens when it breaks, who can close it, and whether the correct answer is simply do nothing because the existing interface already works.
That is a better test than searching for examples that always confirm my opening intuition.
The bowl is the easy part
Suppose the answer at one location really is that a dog-level endpoint would be useful.
The hardware is not difficult to imagine.
Put a shallow basin lower down.
But the moment the bowl becomes public infrastructure, it stops being just a bowl.
Mouths may touch it. Paws and dirt can reach it. Water can remain in it. Drainage can block. Cleaning becomes part of the operation. Frost matters. If it connects to a potable network, the design also has to address backflow and source protection. If the animal component breaks, somebody has to decide whether only that component closes or the whole fountain closes with it.
The problem is not simply standing water versus fresh water.
Fresh water can enter a dirty interface.
A pond can hold water for months and be functioning exactly as intended.
A dog basin may need water to remain long enough for the animal to drink, then drain rather than persist as a permanently shared reservoir.
So the design variable is not just whether water stands still.
It is whether its residence time matches the biological function and the operational model.
And the correct result of an infrastructure investigation must sometimes be:
do nothing.
Otherwise the investigation is only a machine for inventing hardware.
The exposure changes the water
Then there is the source.
If a city deliberately provides water for drinking, controlled drinking-water quality is the conservative source candidate. That is the conservative starting point if a dog interface branches from an existing human drinking-water fountain.
But carry that logic twenty metres to a tree and it stops making sense.
A tree trench may be designed to catch rain or runoff and keep some of it in the root zone instead of exporting it immediately into a drain.
A rain garden needs temporary retention.
A pond needs persistent surface water.
Reclaimed water may be appropriate for one controlled use and inappropriate for another.
So I cannot treat “water” as one interchangeable input either.
The exposure defines the water quality.
Again, this is ordinary fit-for-purpose engineering before it is anything else.
The useful question is whether the category changes somewhere near the interface and nobody notices because upstream everything is still called water infrastructure.
The right water can meet the wrong body.
The right body can meet the wrong interface.
The right interface can be fed by the wrong source.
Someone has to own the last metre
The physical system eventually becomes an institutional one.
Paris is useful here because the public drinking-fountain chain is comparatively readable in the material I have found so far.
For this asset class, the City chooses fountain locations, while Eau de Paris maintains the fountain network and controls its water quality.
The point is not that one institution owns every water object.
For this investigation, I treat a public interface as more legible when somebody can answer operational questions about the specific asset.
And this is where the article from the day before comes back.
Writing When the Dog Outside the Supermarket Becomes an Infrastructure Problem, I had already arrived at a sharper principle:
Control makes responsibility possible; foreseeable reliance defines its scope.
The fountain did not produce a second responsibility rule. It gave the same rule another object.
If a city provides only a human drinking point and the owner supplies the bowl, much of the dog’s last metre remains mediated by the owner.
If the city deliberately adds a dog basin, chooses how it fills and drains, sets the cleaning cycle, controls shutdown and reopening, and invites repeated public use, then it has taken control of more of that interaction. People may also begin to rely on the interface being there and working as intended.
That does not by itself settle legal liability.
It tells me where to look for operational responsibility.
Who controls this interface, at this location, in this state?
Who inspects it?
Who cleans it?
Who receives the failure report?
Who can isolate it?
Who decides it is ready to reopen?
A functioning pipe does not guarantee a functioning endpoint.
A fountain can exist on a map and still be closed.
A beautifully designed dog basin can become useless if nobody owns its cleaning cycle.
The missing component may not be another piece of stainless steel.
It may be a responsibility field in the asset register.
The waterhole did not survive intact
I liked the sentence when I first thought of it:
We did not remove the animals from the city. We removed the waterhole.
The research did not let me keep it as a factual claim.
In some places, urbanisation has removed ponds, buried streams or reduced surface-water area.
In others, cities have created or restored ponds, wetlands, rain gardens, tree trenches and other blue-green systems.
There is no single historical waterhole that the modern city simply deleted.
The title survives for me only as a metaphor for a missing or mismatched interface.
And even that may still be too strong.
But the smaller question has survived every attempt I made to kill it.
For a particular place and a particular living function:
- Is water present?
- Is it actually usable?
- Does the source fit the exposure?
- Who controls the interface, and what reliance does that control invite?
Those questions do not force one answer.
For the dog outside the library, the answer may still be:
bring a bowl.
At Tino Rossi, the answer is currently:
the recovered public record tells me the dog and the water are both there, but not how they meet.
In Washington Square Park, the municipal project record says the dog was deliberately included in the water interface.
For a bat, the intervention may be geometry rather than more water.
For a tree, it may be letting rain remain where roots can use it.
For a public fountain, perhaps the hardware is fine and the remaining problem is institutional: nobody can tell who owns the failure.
That difference is the point.
Looking down
I only noticed the fountain because the dog-hook article from the day before was still sitting in my head.
That is probably why this investigation began with a dog rather than a drainage map or a biodiversity plan.
But it did not stay a dog article.
The dog was just the thing that made me look down.
Paris had already taught me to look at the visible fountain and ask what invisible system had to exist behind it.
The dog-hook article had made me look at control, reliance and responsibility at the final interface.
This time those two lines of thought met at the same object.
What happens after the pipe works?
Who can actually meet the water?
Under what conditions?
And when the city chooses to design that final relationship, how much of it does the city now control — and how much reliance does that control invite?
The bubbles went up.
The question was what happened below them.