Helen Hou-Sandi 3adcda996b Improve oEmbed caching. Introduces the concept of a TTL for oEmbed caches and a filter for oembed_ttl.
We will no longer replace previously valid oEmbed responses with an `{{unknown}}` cache value. When this happens due to reaching a rate limit or a service going down, it is data loss, and is not acceptable. This means that oEmbed caches for a post are no longer deleted indiscriminately every time that post is saved.

oEmbed continues to be cached in post meta, with the addition of a separate meta key containing the timestamp of the last retrieval, which is used to avoid re-requesting a recently cached oEmbed response. By default, we consider a valued cached in the past day to be fresh. This can greatly reduce the number of outbound requests, especially in cases where a post containing multiple embeds is saved frequently.

The TTL used to determine whether or not to request a response can be filtered using `oembed_ttl`, thus allowing for the possibility of respecting the optional oEmbed response parameter `cache_age` or altering the period of time a cached value is considered to be fresh.

Now that oEmbeds are previewed in the visual editor as well as the media modal, oEmbed caches are often populated before a post is saved or published. By pre-populating and avoiding having to re-request that response, we also greatly reduce the chances of a stampede happening when a published post is visible before oEmbed caching is complete.

As it previously stood, a stampede was extremely likely to happen, as the AJAX caching was only triggered when `$_GET['message']` was 1. The published message is 6. We now trigger the caching every time `$_GET['message']` is present on the edit screen, as we are able to avoid triggering so many HTTP requests overall.

props markjaquith. fixes #14759. see #17210.


git-svn-id: https://develop.svn.wordpress.org/trunk@28972 602fd350-edb4-49c9-b593-d223f7449a82
2014-07-03 16:44:15 +00:00
Description
No description provided
243 MiB
Languages
PHP 80.5%
CSS 9.4%
JavaScript 8.5%
SCSS 0.9%
HTML 0.7%