如何在App Engine上实现轻量级的发布订阅服务？
在他的 Google I/O 2009“App Engine 离线处理：展望”演讲中（视频< /a>、幻灯片），Brett Slatkin 介绍任务队列服务。
他指出
发布-订阅系统最大化交易，解耦：
- 每秒处理大量小事务
- 一对多扇出并更换接收器
- 保证排序、过滤、两阶段提交
并且特别强调
我们的新 API 实现了队列，而不是发布-订阅
我只对这些功能的一个子集感兴趣：
更改选定/固定内部的一对多扇出接收处理程序保证排序、过滤、两阶段
提交 目标是简化同一 Web 应用程序的不同模块之间通知/消息的发布。示例使用场景如下：
- 使支付模块了解账单的接收。
- 使用户能够跟踪他决定关注/加注星标的特定域对象的更改。
在任务队列服务之上实现这些的正确方法是什么？
During his Google I/O 2009 "Offline processing on App Engine: A look ahead" presentation (video, slides), Brett Slatkin presents the Task Queue service.
He states that
Pub-sub systems maximize transactions, decoupling:
- Large numbers of small transactions per second
- One-to-many fan-out with changing receivers
- Guaranteed ordering, filtering, two-phase commit
And specifically emphasises
Our new API implements queueing, not pub-sub
I'm interested in only a subset of those functionalities:
- One-to-many fan-out with
changingselected/fixed internal receiver handlers Guaranteed ordering, filtering, two-phase commit
Targeted goal would be to ease publishing of notifications/messages between different modules of the same web application. Sample usage scenarios cases would be:
- Making the payment module aware of receivals of bills.
- Making a user able to track changes of a particular domain object he has decided to follow/star.
What would be the correct way to implement these on top of the Task Queue service ?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
考虑使用Cloud Pub/Sub。它是内部 Pub/Sub 技术的外部化版本，该技术已在 Google 内部广泛使用了大约 7 年，并被证明具有可扩展性和稳健性。截至今天，它仍处于测试阶段，但我们很快就会推出 GA（几个月后）。
Consider using Cloud Pub/Sub. It's the externalized version of the internal Pub/Sub technology, which has been used extensively within google for about 7 years and proven to be scalable and robust. As of today it's still beta, but we're going to GA soon (in few months).
Google 提供了一个在 AppEngine 上运行的 PubSubHubbub 实现示例。您可以看看他们是如何做的，或者删除代码，直到它满足您的需求。
There is an example implementation of PubSubHubbub from Google that runs on AppEngine. You can have a look at how they do it or remove code until it fits your needs.