Cache invalidation

[henvic]

Cache invalidation

[eduardolundgren]

Either solution SGTM. Worth researching with one you think makes more sense, run some tests and take the decision. Let me know.

[henvic]

Here are my findings: the best way to deal with caching for the RequestScreen is to not try to handle it at all. Just let the browser / server do it using ETags or whatever caching policy it's used. The developers just have to follow the standards (on the HTTP level, not HTML Pragma / Cache-Content meta tags).

[FagnerMartinsBrack]

I assumed that leveraging http default cache machanism was the default behavior.

I am curious, what was the original motivation for using an alternative cache method?

[henvic]

Senna is a spin-off of AlloyUI's surface module, which is used by the Liferay portal.

The reason why a JS-based caching mechanism was implemented there was because we needed it for some custom screens on the Liferay portal project.

How come this JS-based caching mechanism might be useful? Imagine you've some custom screens that are generated on the client-side or receive POST* request data. A JS-based caching mechanism is the way to go to improve performance.

• Bad HTTP verb choice, for sure.

[FagnerMartinsBrack]

Yeah, it makes sense.
For "senna" standalone project though, the POST request case is something to be considered if #59 is ever implemented. In that case it is a valid http verb.

[henvic]

This doesn't relate with #59 directly.

A PostScreen could, theoretically, make use of it, but it's an abuse. We had this special case where caching it helped us achieve our goals due to legacy code that is going to be rewritten at some point (and was risky to change right now).

POST requests MUST NOT be cached, unless the response includes Cache-Control or Expires headers, which in this case will be automatically resolved by the browser and Senna MUST NOT deal with it.

References:
9 Method Definitions
13 Caching in HTTP

[FagnerMartinsBrack]

Sure, standards should be respected.

I mentioned that it could be considered because there is is some evidence that most browsers were not that consistent in handling POST request caching. But that was long ago and this is probably not an issue anymore, haven't tested it.

[eduardolundgren]

This should be enough for now https://github.com/liferay/senna.js/blob/master/src/app/App.js#L241. Thanks!