Adress documentation comment

docs/examples.multiple_connections.rst

@@ -96,23 +96,46 @@ Production Configuration
 
 For production use, consider these settings:
 
+**RetryConfig Parameters**
+
+The `RetryConfig` class controls connection retry behavior with exponential backoff:
+
+- **max_retries**: Maximum number of retry attempts before giving up (default: 3)
+- **initial_delay**: Initial delay in seconds before the first retry (default: 0.1s)
+- **max_delay**: Maximum delay cap to prevent excessive wait times (default: 30.0s)
+- **backoff_multiplier**: Exponential backoff multiplier - each retry multiplies delay by this factor (default: 2.0)
+- **jitter_factor**: Random jitter (0.0-1.0) to prevent synchronized retries (default: 0.1)
+
+**ConnectionConfig Parameters**
+
+The `ConnectionConfig` class manages multi-connection behavior:
+
+- **max_connections_per_peer**: Maximum connections allowed to a single peer (default: 3)
+- **connection_timeout**: Timeout for establishing new connections in seconds (default: 30.0s)
+- **load_balancing_strategy**: Strategy for distributing streams ("round_robin" or "least_loaded")
+
+**Load Balancing Strategies Explained**
+
+- **round_robin**: Cycles through connections in order, distributing load evenly. Simple and predictable.
+- **least_loaded**: Selects the connection with the fewest active streams. Better for performance but more complex.
+
 .. code-block:: python
 
     from libp2p.network.swarm import ConnectionConfig, RetryConfig
 
     # Production-ready configuration
     retry_config = RetryConfig(
-        max_retries=3,
-        initial_delay=0.1,
-        max_delay=30.0,
-        backoff_multiplier=2.0,
-        jitter_factor=0.1
+        max_retries=3,           # Maximum retry attempts before giving up
+        initial_delay=0.1,       # Start with 100ms delay
+        max_delay=30.0,          # Cap exponential backoff at 30 seconds
+        backoff_multiplier=2.0,  # Double delay each retry (100ms -> 200ms -> 400ms)
+        jitter_factor=0.1        # Add 10% random jitter to prevent thundering herd
     )
 
     connection_config = ConnectionConfig(
-        max_connections_per_peer=3,  # Balance performance and resources
-        connection_timeout=30.0,     # Reasonable timeout
-        load_balancing_strategy="round_robin"  # Predictable behavior
+        max_connections_per_peer=3,  # Allow up to 3 connections per peer
+        connection_timeout=30.0,     # 30 second timeout for new connections
+        load_balancing_strategy="round_robin"  # Simple, predictable load distribution
     )
 
     swarm = new_swarm(
@@ -120,6 +143,44 @@ For production use, consider these settings:
         connection_config=connection_config
     )
 
+**How RetryConfig Works in Practice**
+
+With the configuration above, connection retries follow this pattern:
+
+1. **Attempt 1**: Immediate connection attempt
+2. **Attempt 2**: Wait 100ms ± 10ms jitter, then retry
+3. **Attempt 3**: Wait 200ms ± 20ms jitter, then retry
+4. **Attempt 4**: Wait 400ms ± 40ms jitter, then retry
+5. **Attempt 5**: Wait 800ms ± 80ms jitter, then retry
+6. **Attempt 6**: Wait 1.6s ± 160ms jitter, then retry
+7. **Attempt 7**: Wait 3.2s ± 320ms jitter, then retry
+8. **Attempt 8**: Wait 6.4s ± 640ms jitter, then retry
+9. **Attempt 9**: Wait 12.8s ± 1.28s jitter, then retry
+10. **Attempt 10**: Wait 25.6s ± 2.56s jitter, then retry
+11. **Attempt 11**: Wait 30.0s (capped) ± 3.0s jitter, then retry
+12. **Attempt 12**: Wait 30.0s (capped) ± 3.0s jitter, then retry
+13. **Give up**: After 12 retries (3 initial + 9 retries), connection fails
+
+The jitter prevents multiple clients from retrying simultaneously, reducing server load.
+
+**Parameter Tuning Guidelines**
+
+**For Development/Testing:**
+- Use lower `max_retries` (1-2) and shorter delays for faster feedback
+- Example: `RetryConfig(max_retries=2, initial_delay=0.01, max_delay=0.1)`
+
+**For Production:**
+- Use moderate `max_retries` (3-5) with reasonable delays for reliability
+- Example: `RetryConfig(max_retries=5, initial_delay=0.1, max_delay=60.0)`
+
+**For High-Latency Networks:**
+- Use higher `max_retries` (5-10) with longer delays
+- Example: `RetryConfig(max_retries=8, initial_delay=0.5, max_delay=120.0)`
+
+**For Load Balancing:**
+- Use `round_robin` for simple, predictable behavior
+- Use `least_loaded` when you need optimal performance and can handle complexity
+
 Architecture
 ------------
 

libp2p/network/swarm.py

@@ -63,7 +63,26 @@ logger = logging.getLogger("libp2p.network.swarm")
 
 @dataclass
 class RetryConfig:
-    """Configuration for retry logic with exponential backoff."""
+    """
+    Configuration for retry logic with exponential backoff.
+
+    This configuration controls how connection attempts are retried when they fail.
+    The retry mechanism uses exponential backoff with jitter to prevent thundering
+    herd problems in distributed systems.
+
+    Attributes:
+        max_retries: Maximum number of retry attempts before giving up.
+                     Default: 3 attempts
+        initial_delay: Initial delay in seconds before the first retry.
+                      Default: 0.1 seconds (100ms)
+        max_delay: Maximum delay cap in seconds to prevent excessive wait times.
+                  Default: 30.0 seconds
+        backoff_multiplier: Multiplier for exponential backoff (each retry multiplies
+                           the delay by this factor). Default: 2.0 (doubles each time)
+        jitter_factor: Random jitter factor (0.0-1.0) to add randomness to delays
+                      and prevent synchronized retries. Default: 0.1 (10% jitter)
+
+    """
 
     max_retries: int = 3
     initial_delay: float = 0.1
@@ -74,7 +93,21 @@ class RetryConfig:
 
 @dataclass
 class ConnectionConfig:
-    """Configuration for multi-connection support."""
+    """
+    Configuration for multi-connection support.
+
+    This configuration controls how multiple connections per peer are managed,
+    including connection limits, timeouts, and load balancing strategies.
+
+    Attributes:
+        max_connections_per_peer: Maximum number of connections allowed to a single
+                                 peer. Default: 3 connections
+        connection_timeout: Timeout in seconds for establishing new connections.
+                           Default: 30.0 seconds
+        load_balancing_strategy: Strategy for distributing streams across connections.
+                                Options: "round_robin" (default) or "least_loaded"
+
+    """
 
     max_connections_per_peer: int = 3
     connection_timeout: float = 30.0