Selenium WebDriver operates through a three-tier architecture analogous to a taxi service:

• The test automation script acts as the passenger, specifying navigation and actions.
• Browser drivers function as the taxi drivers, interpreting commands and controlling browser behavior.
• Web browsers serve as the vehicles, executing actual user interface operations.

At the technical level, these components correspond to:

1. WebDriver APIs available in multiple languages (Java, Python, C#).
2. Browser-specific executables (chromedriver.exe, geckodriver.exe, IEDriverServer.exe).
3. Standard web browsers like Chrome, Firefox, and Internet Explorer.

Communication occurs via HTTP requests between the test script and the browser driver server. Each command from the automation script is packaged into an HTTP request sent to the driver's embedded HTTP server, which then translates and forwards instructions to the browser.

The underlying protocol used is JSON Wire Protocol, which defines standardized JSON structures for request/response bodies over HTTP. Common HTTP methods include:

• GET: Retrieve information such as page titles.
• POST: Execute actions like element location or click operations.

Response status codes provide specific feedback:

• Status 7: NoSuchElement
• Status 11: ElementNotVisible
• Status 200: Success

To illustrate, starting chromedriver initializes a server on port 9515:

chromedriver
Starting ChromeDriver 2.39.562713 (dd642283e958a93ebf6891600db055f1f1b4f3b2) on port 9515
Only local connections are allowed.

A typical session initiation involves sending a POST request to http://localhost:9515/session with capability specifications:

import requests
import json

session_url = 'http://localhost:9515/session'
session_data = {
    "capabilities": {
        "firstMatch": [{}],
        "alwaysMatch": {
            "browserName": "chrome",
            "platformName": "any"
        }
    },
    "desiredCapabilities": {
        "browserName": "chrome",
        "platform": "ANY"
    }
}
r_session = requests.post(session_url, json=session_data)
print(json.dumps(r_session.json(), indent=2))

Following session creation, subsequent operations target specific endpoints:

Opening a URL:

url = 'http://localhost:9515/session/{}/url'.format(session_id)
payload = {"url": "https://www.baidu.com"}
requests.post(url, json=payload)

Finding elements:

url = 'http://localhost:9515/session/{}/element'.format(session_id)
payload = {"using": "css selector", "value": ".postTitle a"}
requests.post(url, json=payload)

Interacting with elements:

url = 'http://localhost:9515/session/{}/element/{}/click'.format(session_id, element_id)
payload = {"id": element_id}
requests.post(url, json=payload)

Complete example demonstrating basic browser control:

import requests
import time

capabilities = {
    "capabilities": {
        "alwaysMatch": {
            "browserName": "chrome"
        },
        "firstMatch": [{}]
    },
    "desiredCapabilities": {
        "platform": "ANY",
        "browserName": "chrome",
        "version": "",
        "chromeOptions": {
            "args": [],
            "extensions": []
        }
    }
}

# Launch browser
res = requests.post('http://127.0.0.1:9515/session', json=capabilities).json()
session_id = res['sessionId']

# Navigate to page
requests.post('http://127.0.0.1:9515/session/{}/url'.format(session_id),
             json={"url": "http://www.baidu.com"})
time.sleep(3)

# Close session
requests.delete('http://127.0.0.1:9515/session/{}'.format(session_id))

Understanding these fundamentals aids debugging and provides insight into low-level browser automation mechanics.