Opened 5 years ago
Last modified 5 years ago
#49627 new defect (bug)
oEmbedding - Privacy by design - Inconsistencies & Legal concerns
Reported by: | arena | Owned by: | |
---|---|---|---|
Milestone: | Awaiting Review | Priority: | normal |
Severity: | normal | Version: | |
Component: | Embeds | Keywords: | |
Focuses: | ui, javascript, rest-api, privacy | Cc: |
Description
oEmbedding is now driven by two files :
wp-includes/class-wp-oembed.php
and its list of "popular" oEmbed providers : array ( pattern => array( oembed_url, regex [boolean] ) )
This list is loaded in wp-settings.php
with this instruction : $GLOBALS['wp_embed'] = new WP_Embed();
This list of providers is filterable with hook 'oembed_providers'
, but the hook appears to be fired before 'plugins_loaded'
and 'init'
hooks.
The modus operandi to alter the "popular" providers list is described here : https://wordpress.org/support/article/embeds/ .
wp-includes/js/dist/block-library.js
The data structure is here different due to some visual aspects (Gutenberg)
Several variables with name like 'embedXxxxxIcon'
are objects containing svg icons
and another list of oEmbed providers with a specific variable called 'common' which is a set of
-- name => settings ( title, name svg icon, keywords, description )
=> patterns
1) the lists of oEmbed providers are not identical in php and javascript.
2) in each list, the patterns for each oEmbed provider can be different.
3) the once driven oEmbedding process by the php array that could be filtered or altered ("popular" oEmbed providers) is not possible anymore.
If you are using the Classic editor or Gutenberg, the process to retrieve the data and the structure of the data are different as shown in screenshot1 (Classic trunk) and screenshot2 (Gutenberg 5.3).
'Classic editor' is using admin-ajax.php
and retrieves a specific json format.
'Gutenberg' is using the rest-api and calls a url such as 'index.php?rest_route=/oembed/1.0/proxy&url....&_locale=user'
and retrieves the full oembed json from the provider.
In the class-wp-oembed-controller, the code executed seems similar to Classic editor (data2html) but the result is obviously different.
Both are caching the html code in post_meta table, 'Gutenberg' is caching the json result of oembed in options table.
This means that the filter 'oembed_result' is by-passed when oembed content is displayed in 'Gutenberg'.
4) if a site decides to remove one or all references to "popular" or "trusted" (vocabulary differs in code comments) oEmbed providers or add its own, this is not possible anymore when using Gutenberg, because the javascript providers list is hard-coded and cannot be altered.
For obvious local regulations, it is important to be consistent and coherent in the oembedding policy choices of each site for each site.
This is a regression and an infringement to the "Privacy by design" policy adopted by wordpress open source project.
5) i am not going to discuss how the "popular" providers got elected by a "popular" vote or an electoral college and if they had a superpac to get elected (or more accurately "be referenced").
On some websites, these arbitrary choices that cannot be altered may contervene to website policy e.g. educational blogs for kids, or local laws.
In some countries, this can be considered as political [ethymology definition : "life in the city"] choice and can put at risk wordpress users : loading in their browsers references to forbidden sites.
Surveillance is everywhere and not all countries are democracies.
6) it is important, that in WordPress, all "popular" oEmbed providers removed or added by filter, be removed or added in any form of code (Gutenbeg specific blocks included) (privacy by design).
This implies a unique reference list of providers (filterable/alterable in php) with some extra information for visual aspects under Gutenberg.
The javascript variables must be generated (or not if providers is an empty array) from the php list (with some default values if extra information is not provided).
To simply and easily :
a) add a new provider on both sides and specific svg icons.
b) remove a "popular" provider with an easy
wp_oembed_remove_provider( 'tactic' );
rather than
wp_oembed_remove_provider( '#https?://(www\.)?tactic\.com/.*/video/.*#i' );
c)
wp_oembed_remove_all_providers()
is also welcome.
7) it could be interesting to offer optional and intermediate solutions to oEmbedding as it is processed today (privacy by design) :
a) insert specific html ( usually iframe : fully intrusive)
b) insert link to url with thumbnail when available<a href=...><img src=thumbnail_url ... /></a>
(less intrusive)
c) insert link as text<a href=url ...>(some title or text if available or url)</a>
(no oEmbedding : not intrusive)
Requires to cache the complete json answer indifferently ('Classic' or 'Gutenberg').
8) it could be interesting to encapsulate the oembedding result in a custom html5 element (privacy by design)
such as <wp-oembed> ... </wp-oembed>
to allow theme developpers to add specific js on embedded content.
9) The oEmbed providers listed in wordpress are referenced by wordpress. Definitely change the vocabulary used in the code comments.
Do not use "popular" or "trusted" providers but 'referenced providers', because this is the most accurate definition.
Regards
related #43713,
related https://wordpress.org/support/article/embeds/
ps : as a remainder, GDPR chapter IV title is Controller and Processor.
Attachments (2)
Change History (6)
#1
in reply to:
↑ description
@
5 years ago
Replying to arena:
oEmbedding is now driven by two files :
wp-includes/class-wp-oembed.php
and its list of "popular" oEmbed providers : array ( pattern => array( oembed_url, regex [boolean] ) )This list is loaded in
wp-settings.php
with this instruction :$GLOBALS['wp_embed'] = new WP_Embed();
This list of providers is filterable with hook'oembed_providers'
, but the hook appears to be fired before'plugins_loaded'
and'init'
hooks.
Note the difference between WP_Embed
and WP_oEmbed
.
The former is indeed created early in wp-settings.php
, but that does not create the WP_oEmbed
object.
The latter is created in _wp_oembed_get_object()
after plugins are loaded. wp_oembed_add_provider()
and wp_oembed_remove_provider()
specifically check for did_action( 'plugins_loaded' )
.
#3
@
5 years ago
@SergeyBiryukov thank you for being more precise than i was.
I hope the "Privacy by design" issue will be addressed
screenshot1 classic trunk (5.5-alpha-47436)