Improve caching a bit more

[marcoscaceres]

Emailed with mnot briefly, he suggested:

"You can drop Expires and just do CC: max-age."

I'm wondering if we could play with extending the cache duration for up to a week? If it's still going to revalidate against the hashed content (via etag), then it should be fairly safe to extend to a week.

Anyone have opinions?

[tobie]

Can you explain what that would change?

[darobin]

Sending both Expires and Cache-Control: max-age shouldn't be necessary unless you intend to target old HTTP/1.0 clients. You can drop the former.

My understanding is that if you have an ETag then you can use expiration dates farther in the future, because the client will use If-Match to which you can respond with a 304, whereas after the expiration time it will make an unconditional request (disclaimer: I'm no expert on this, it's just my understanding). Obviously this is only useful if you handle the conditions.

[tobie]

Oh, so we hit the server everytime no matter what?

[darobin]

You're probably not going to hit the server every time. But sometimes you are. The behaviour on reload for instance (and spec dev is all about the reloads) is browser-dependent. You're going to get If-Match requests after the cache expiration time too; it's not a bad idea to honour them if it's cheap enough for you.