---
title: "Weekly Technical Report for the Second Week of February 2023"
url: https://blog.bktus.com/en/archives/lbe9e3/
author: "Saturneric"
date: 2023-06-21
language: en-US
categories: ["Feelings"]
tags: ["weeklies","Tencent","Legacy Systems","Redis"]
translation_zh: https://blog.bktus.com/archives/q1pqco/
---

# Weekly Technical Report for the Second Week of February 2023

From the end of January to early February, the Chinese New Year period required
on-call support to ensure system stability. During this time, the best possible
outcome is simple: no incidents. Fortunately, everything remained stable, and no
major issues occurred.

This week, I mainly evaluated the impact of a new business requirement. One
important lesson I have learned is that, when working on complex
systems—especially those you are not fully familiar with—the safest approach is
to minimize changes as much as possible.

This is not about being overly conservative, but about controlling uncertainty.
In large systems, there are often hidden dependencies and non-obvious logic
paths. Making broad or intrusive changes without a deep understanding can easily
introduce subtle issues that may only surface much later.

This insight comes from experience. By the time I wrote this, I had already
encountered similar situations more than once. In one case, I made a significant
modification to a service. Everything appeared to work correctly during testing
and shortly after deployment. However, weeks later, I discovered that the change
had unintentionally affected a chain of upstream and downstream interactions.

In systems that have evolved over time and passed through multiple hands, there
is always hidden complexity. Some mechanisms may not be immediately visible but
are critical to the overall system behavior. As a general rule: avoid modifying
core logic or frameworks unless absolutely necessary.

In addition to this evaluation work, I also resolved an issue related to a
security encryption service that was not compatible with Chinese characters.

The root cause was that the service directly used plaintext input as Redis keys.
When the input contained Chinese characters, the keys could be stored
successfully but could not be reliably retrieved due to encoding
inconsistencies.

To fix this, I changed the key generation strategy. Instead of using the
original plaintext, I applied a hash function to generate the Redis key. This
approach has several advantages:

- avoids encoding and compatibility issues
- ensures consistent key lookup
- improves security by not exposing raw business data in keys

This change made the system more robust and predictable.
