This is an automated email from the ASF dual-hosted git repository.
corgy pushed a commit to branch dev
in repository https://gitbox.apache.org/repos/asf/seatunnel.git
The following commit(s) were added to refs/heads/dev by this push:
new 12223d510f [Docs] Remove future directions section from design
philosophy (#10556)
12223d510f is described below
commit 12223d510f436171ee8effa841f47b88de706290
Author: David Zollo <[email protected]>
AuthorDate: Thu Mar 5 10:13:47 2026 +0800
[Docs] Remove future directions section from design philosophy (#10556)
---
docs/en/architecture/design-philosophy.md | 28 ++++++----------------------
docs/zh/architecture/design-philosophy.md | 28 ++++++----------------------
2 files changed, 12 insertions(+), 44 deletions(-)
diff --git a/docs/en/architecture/design-philosophy.md
b/docs/en/architecture/design-philosophy.md
index ff4a1f3d54..98f13612c3 100644
--- a/docs/en/architecture/design-philosophy.md
+++ b/docs/en/architecture/design-philosophy.md
@@ -445,46 +445,30 @@ Many sinks only need Writer + Committer;
AggregatedCommitter is for complex case
However, for Flink translation, SeaTunnel checkpoints align with Flink
checkpoints to avoid duplication.
-## 6. Future Directions
+## 6. Lessons Learned
-### 6.1 Planned Enhancements
-
-- **Dynamic Scaling**: Add/remove workers during job execution
-- **Adaptive Batch Size**: Auto-tune batch sizes based on throughput
-- **Query Pushdown**: Push filters/projections to sources
-- **Vectorized Execution**: Process batches of rows (columnar)
-- **Speculative Execution**: Mitigate stragglers
-
-### 6.2 Research Directions
-
-- **Machine Learning Integration**: ML-based optimization (split sizing,
parallelism)
-- **Unified Batch and Streaming**: True unified processing model
-- **Global Query Optimization**: Cross-pipeline optimization
-
-## 7. Lessons Learned
-
-### 7.1 What Worked Well
+### 6.1 What Worked Well
1. **Engine Independence**: Validated by successful Zeta engine addition
without API changes
2. **Split-based Parallelism**: Scales well to 1000+ parallel tasks
3. **Explicit Schema**: Caught many bugs early, enabled schema evolution
4. **Two-Phase Commit**: Reliable exactly-once semantics
-### 7.2 What Could Be Better
+### 6.2 What Could Be Better
1. **API Complexity**: Enumerator/Committer adds learning curve for simple
connectors
2. **Class Loader Issues**: Occasional conflicts with shaded dependencies
3. **Checkpoint Latency**: Large state causes checkpoint delays
4. **Documentation Gaps**: Architecture docs lagged behind code
-### 7.3 If Starting Over
+### 6.3 If Starting Over
1. **Simplify API**: Provide higher-level abstractions for simple sources/sinks
2. **Async I/O Support**: First-class async API for non-blocking connectors
3. **Built-in Metrics**: Standardized metrics collection in API
4. **Schema Registry Integration**: Tighter integration with external schema
registries
-## 8. Conclusion
+## 7. Conclusion
SeaTunnel's architecture reflects careful trade-offs between competing
concerns:
- Engine independence vs engine-specific optimization
@@ -494,7 +478,7 @@ SeaTunnel's architecture reflects careful trade-offs
between competing concerns:
The V2 redesign addressed major V1 limitations while establishing principles
for long-term evolution. Understanding these design philosophies helps
contributors make consistent decisions and users understand SeaTunnel's
strengths and appropriate use cases.
-## 9. References
+## 8. References
- [Architecture Overview](overview.md)
- [Source Architecture](api-design/source-architecture.md)
diff --git a/docs/zh/architecture/design-philosophy.md
b/docs/zh/architecture/design-philosophy.md
index e37445f5cc..d012f48f74 100644
--- a/docs/zh/architecture/design-philosophy.md
+++ b/docs/zh/architecture/design-philosophy.md
@@ -412,46 +412,30 @@ V2 是未来；V1 处于维护模式。
但是，对于 Flink 转换，SeaTunnel 检查点与 Flink 检查点对齐以避免重复。
-## 6. 未来方向
+## 6. 经验教训
-### 6.1 计划增强
-
-- **动态扩缩容**：在作业执行期间添加/移除工作节点
-- **自适应批量大小**：根据吞吐量自动调整批量大小
-- **查询下推**：将过滤器/投影下推到数据源
-- **向量化执行**：处理批量行（列式）
-- **推测执行**：缓解掉队者
-
-### 6.2 研究方向
-
-- **机器学习集成**：基于 ML 的优化（分片大小、并行度）
-- **统一批处理和流处理**：真正的统一处理模型
-- **全局查询优化**：跨管道优化
-
-## 7. 经验教训
-
-### 7.1 成功之处
+### 6.1 成功之处
1. **引擎独立性**：通过成功添加 Zeta 引擎而无需 API 更改得到验证
2. **基于分片的并行度**：扩展到 1000+ 并行任务
3. **显式模式**：尽早捕获许多错误，实现模式演化
4. **两阶段提交**：可靠的精确一次语义
-### 7.2 可以改进之处
+### 6.2 可以改进之处
1. **API 复杂性**：枚举器/提交器增加了简单连接器的学习曲线
2. **类加载器问题**：遮蔽依赖偶尔冲突
3. **检查点延迟**：大状态导致检查点延迟
4. **文档差距**：架构文档落后于代码
-### 7.3 如果重新开始
+### 6.3 如果重新开始
1. **简化 API**：为简单的数据源/Sink 提供更高级的抽象
2. **异步 I/O 支持**：非阻塞连接器的一等异步 API
3. **内置指标**：API 中的标准化指标收集
4. **模式注册表集成**：与外部模式注册表更紧密的集成
-## 8. 结论
+## 7. 结论
SeaTunnel 的架构反映了竞争关注点之间的仔细权衡：
- 引擎独立性 vs 引擎特定优化
@@ -461,7 +445,7 @@ SeaTunnel 的架构反映了竞争关注点之间的仔细权衡：
V2 重新设计解决了 V1 的主要局限性，同时建立了长期演进的原则。理解这些设计理念有助于贡献者做出一致的决策，并帮助用户了解 SeaTunnel
的优势和适用场景。
-## 9. 参考资料
+## 8. 参考资料
- [架构概览](overview.md)
- [数据 Source 架构](api-design/source-architecture.md)
