Open edX官方提供5种方式扩展edx

github地址：https://github.com/edx/edx-platform/wiki/Five-ways-to-extend-edX

1. jsinput -- 创建介绍js级别的组件并提供getState(), setState(), 和 getGrade()方法
2. LTI -- edX 现在支持 LTI 1.1 , LTI 2.0 还是开发中
3. custom grader -- Code can be run on an external server to do arbitrary work to grade problems. With our Database Class we've had some success returning not just grades, but also an HTML block to be rendered with an answer. We return the query result, complete with HTML table formatting tags. See this screenshot as an example: the contents of the blue box is what was returned by the grader.
4. XBlock
5. hack on core code

	JSinput	LTI	External Grader	XBlock	Hack The Core!
开发成本	Low	Low	Med	Med	High
语言	JS	any	any	Python	Python
开发环境	no	no	yes	yes	yes
本地组件	no	yes	yes	no	no
需要edx整合	no	no	yes	yes	yes
基于iframe的ui	yes	no [1]	yes	yes	yes
移动端	maybe	maybe	yes	yes	yes
服务端分级	no [2]	yes	yes	yes	yes
日志记录	no [3]	no	limited	yes (?)	yes
Studio里的约束	no	no	no	yes	no
隐私相关	no	possibly [4]	possibly [4]	no	no

Notes:

• [1] Only LTI components delivered via https can be iframed in Many are served over http only. And even then they usually have their own look and feel. For example, Piazza can be iFramed in, but has its own navigation elements and their color scheme (seehttp://networking.class.stanford.edu/ for an example).
• [2] JSInput does have a small place where python could do server side grading, but doing it there would be pretty hacky and difficult to maintain.
• [3] JSInput really only exposes the getState, putState, getGrade methods. But is there any reason why we can’t also document / publish the tracking endpoint (/events/user_track I believe)
• [4] Privacy and secrecy issues are important in some areas of the world, and only becoming more important because of the news. If you have to self host OpenEdx because of that, this row indicates the guarantees that you can offer your students in that area. "Possibly" indicates that the guarantees will have to depend on the solution chosen for that particular use case. Note that both LTI and external graders allow for anonymization, but this might not be enough to allow for the use of some external tools within the constraints.