//
            // If we were to cancel taskFuture, that would let the next task start while the old
            // one is still running.
            //
            // Now, maybe we could tweak our implementation to not start the next task until the
            // callable actually completes. (We could detect completion in our wrapper
            // `AsyncCallable task`.) However, our contract also promises:

The OpenAPI schema is what powers the two interactive documentation systems included.

And there are dozens of alternatives, all based on OpenAPI. You could easily add any of those alternatives to your application built with **FastAPI**.

You could also use it to generate code automatically, for clients that communicate with your API. For example, frontend, mobile or IoT applications.

* Response payloads.

You would also have **inline errors** for everything.

And whenever you update the backend code, and **regenerate** the frontend, it would have any new *path operations* available as methods, the old ones removed, and any other change would be reflected on the generated code. 🤓

* **Uvicorn**:
    * Will have the best performance, as it doesn't have much extra code apart from the server itself.
    * You wouldn't write an application in Uvicorn directly. That would mean that your code would have to include more or less, at least, all the code provided by Starlette (or **FastAPI**). And if you did that, your final application would have the same overhead as having used a framework and minimizing your app code and bugs.

# OpenAPI Callbacks { #openapi-callbacks }

You could create an API with a *path operation* that could trigger a request to an *external API* created by someone else (probably the same developer that would be *using* your API).

Embedding a Java runtime in the distribution would provide some benefits, such as allowing the Launcher, Daemon and Workers to run on it, removing the prerequisite of an installed Java runtime.
However, this does not fully remove the prerequisite, as the Wrapper itself would still need an installed Java runtime to execute.

But every time we do:

```Python
Settings()
```

a new `Settings` object would be created, and at creation it would read the `.env` file again.

If the dependency function was just like:

```Python
def get_settings():
    return Settings()
```

we would create that object for each request, and we would be reading the `.env` file for each request. ⚠️

#### A "professional" attack { #a-professional-attack }

Of course, the attackers would not try all this by hand, they would write a program to do it, possibly with thousands or millions of tests per second. And they would get just one extra correct letter at a time.

But doing that, in some minutes or hours the attackers would have guessed the correct username and password, with the "help" of our application, just using the time taken to answer.

import jcifs.internal.smb2.rdma.RdmaMemoryRegion;

/**
 * DiSNI memory region implementation.
 *
 * This class would integrate with DiSNI to provide registered
 * memory regions for high-performance RDMA operations.
 *
 * Note: This is a skeleton implementation. A real implementation would
 * require proper DiSNI integration with actual memory registration.
 */
public class DisniMemoryRegion extends RdmaMemoryRegion {

```
.
├── app
│   ├── __init__.py
│   ├── main.py
│   └── test_main.py
```

The file `main.py` would have:

{* ../../docs_src/async_tests/app_a_py39/main.py *}

The file `test_main.py` would have the tests for `main.py`, it could look like this now:

{* ../../docs_src/async_tests/app_a_py39/test_main.py *}

## Run it { #run-it }

You can run your tests as usual via: