Airline NDC APIs: IATA New Distribution Capability Latency, XML vs. JSON Payload Overhead, and Offer Order Migration

## The Engineering Architecture of Modern Airline Distribution

The global aviation sector is executing a generational migration away from 1970s-era EDIFACT messaging protocols toward IATA's New Distribution Capability (NDC) and ONE Order standards. While NDC gives carriers direct control over retailing, ancillary bundling, and dynamic pricing, it introduces massive computational and bandwidth overhead.

### 1. Protocol Performance Comparison in Flight Search

| Protocol Standard | Format | Avg Shopping Payload Size | Server Response Time (P95) | Look-to-Book Ratio Handling |
|:------------------|:------:|:-------------------------:|:--------------------------:|:---------------------------:|
| Legacy EDIFACT | Compact Text | 12 KB - 48 KB | 420ms | 1,000:1 (Pre-Cached) |
| NDC 17.2 | Verbose XML | 450 KB - 2.8 MB | 3,800ms | 150:1 (Heavy Engine Load) |
| NDC 21.3 | Optimized XML | 180 KB - 850 KB | 1,850ms | 400:1 (Aggregator Cache) |
| Modern JSON Gateway | REST / JSON | 65 KB - 220 KB | 620ms | 800:1 (Edge Decompressed) |

### 2. Mitigating Look-to-Book Load with Edge Caching

Modern flight search engines generate over 10,000 search queries for every ticket issued (look-to-book ratio $\ge 10,000:1$). Direct querying of airline reservation systems at this scale triggers rate limiting. Implementing multi-tier edge caching with short TTLs (120-300 seconds) balances price freshness against infrastructure preservation:

$$P(\text{Stale Price}) = 1 - e^{-\lambda_{\text{fare change}} \cdot \text{TTL}}$$