Page MenuHomePhabricator

Add metainformation for interlanguage links (link rel="alternate" hreflang)
Open, LowPublicFeature

Description

Author: wegge

Description:

Add semantic information for language equivalents of the current page, as determined by interlanguage links, with hreflang links in the <head>. See https://www.mediawiki.org/wiki/Special:Code/MediaWiki/46059#c1385 and https://lists.wikimedia.org/pipermail/wikitech-l/2015-February/080661.html for more information on the syntax.

Cf. T6901 (attributes for interwikis, not the same semantic value); T54429 (language converter already tries to do this but has issues); T70490 (language info for machine translation services).


This patch adds xml:lang links, as described in
http://www.w3.org/TR/2004/WD-xhtml2-20040722/mod-meta.html#sec_20.5.2. to all
pages with interlanguage links. A diff against the current (2005-jan-30 09:00
UTC) CVS REL1_4 is attached. The results can be seen in the source for the two
pages http://playwiki.wegge.dk/MediaWiki and
http://wiki.wegge.dk/MediaWiki_customizations

Diff against current code:

  • Programming/MediaWiki/REL1_4/includes/SkinTemplate.php 2005-01-30

09:21:07.976778061 +0100
+++ playwiki/includes/SkinTemplate.php 2005-01-30 09:42:07.552766175 +0100
@@ -328,7 +328,8 @@ class SkinTemplate extends Skin {

			$nt = Title::newFromText( $l );
			$language_urls[] = array('href' => $nt->getFullURL(),
			'text' => ($wgContLang->getLanguageName( $nt->getInterwiki()) !=

''?$wgContLang->getLanguageName( $nt->getInterwiki()) : $l),

  • 'class' => $wgContLang->isRTL() ? 'rtl' : 'ltr');

+ 'class' => $wgContLang->isRTL() ? 'rtl' : 'ltr',
+ 'language_code' => $nt->getInterwiki());

		}
		if(count($language_urls)) {
			$tpl->setRef( 'language_urls', $language_urls);

diff -rup -xCVS Programming/MediaWiki/REL1_4/skins/MonoBook.php
playwiki/skins/MonoBook.php

  • Programming/MediaWiki/REL1_4/skins/MonoBook.php 2005-01-22

22:45:48.000000000 +0100
+++ playwiki/skins/MonoBook.php 2005-01-30 09:51:16.091509526 +0100
@@ -60,7 +60,13 @@ class MonoBookTemplate extends QuickTemp

<?php if($this->data['usercss'   ]) { ?><style type="text/css"><?php      
  $this->html('usercss'   ) ?></style><?php    } ?>
<?php if($this->data['userjs'    ]) { ?><script type="text/javascript"

src="<?php $this->text('userjs' ) ?>"></script><?php } ?>

<?php if($this->data['userjsprev']) { ?><script

type="text/javascript"><?php $this->html('userjsprev') ?></script><?php } ?>

  • </head>

+ <?php if( $this->data['language_urls'] )
{foreach($this->data['language_urls'] as $langlink) { ?>
+ <link title="<?php echo $langlink['text'] ?> version"
+ rel="alternate"
+ xml:lang="<?php echo $langlink['language_code'] ?>"
+ resource="<?php echo $langlink['href'] ?>" />
+ <?php } } ?>
+</head>

<body <?php if($this->data['body_ondblclick']) { ?>ondblclick="<?php

$this->text('body_ondblclick') ?>"<?php } ?>

<?php if($this->data['nsclass'        ]) { ?>class="<?php

$this->text('nsclass') ?>"<?php } ?>>

<div id="globalWrapper">

Version: 1.4.x
Severity: enhancement
URL: http://www.w3.org/TR/2004/WD-xhtml2-20040722/mod-meta.html#sec_20.5.2.

Details

Reference
bz1433
TitleReferenceAuthorSource BranchDest Branch
Run the AppArmor tests via PHPUnitrepos/abstract-wiki/aw-ci-chart!27jforresterT343378main
Temporarily disable AppArmor testsrepos/abstract-wiki/aw-ci-chart!23kindrobotkr/disable-apparmor-testsmain
Customize query in GitLab

Event Timeline

bzimport raised the priority of this task from to Low.Nov 21 2014, 8:12 PM
bzimport set Reference to bz1433.
bzimport added a subscriber: Unknown Object (MLST).

wegge wrote:

The patch as an attachment

attachment Xml_lang_in_MonoBook.diff ignored as obsolete

Created attachment 4130
Patch updated to apply to trunk as of r25921

attachment bug1433.diff ignored as obsolete

Created attachment 4131
Alternate version of patch with HTML 4.01-compatible link info

The original version of the patch uses <link> attributes recommended in an XHTML 2 draft spec, but these don't seem to be seen by existing user agents.

This alternate patch uses lang, hreflang, and href instead of xml:lang and resource attributes, matching HTML 4.01 recommendations.

I can confirm that the language alternates are listed in Seamonkey's navigation bar if activated and by lynx.

Note that this may actually annoy lynx users, as does the existing OpenSearch description link, but that's because lynx is kind of bad about putting these supplementary links in totally the wrong place. :P

This version of the patch also applies proper HTML escaping...

The hardcoded English string 'version' in the title is probably not too super.

attachment bug1433b.diff ignored as obsolete

Rather than hardcoding this into MonoBook.php, this should probably be done in a skin-independent way, pumping entries into the list of header links on OutputPage which will get used by any skin.

Note also bug 4901, for adding lang attribs to the clickable interlanguage links.

Fixed up in r46059, with a completely different patch.

Reopening per comments on r46059.

Would someone like to test the rel="alternate" on the inline <a href>s and confirm whether it seems to work? If working, they should show up in the navigation bar when enabled in SeaMonkey.

Wiki.Melancholie wrote:

SeaMonkey and "Firefox + Site navigation bar extension" do show this, yes.
[More]
L [Other Versions]

Opera doesn't seem to provide [More], nor [Documents]; so Opera doesn't show this (at least not by default).

ayg wrote:

Putting the rel= on the language link certainly seems like the correct thing here. I don't think we need an extra fifty million <link>s just for the sake of Opera.

Wiki.Melancholie wrote:

Note: I tried this at Betawiki (r46385), and there it's *not* in the <a href=""> link but in a <link>.
Will it work properly in href-links too?


By the way: You could turn around title and rel in <link title="..." rel="alternate" (for uniformity with the other <link> tags).

ayg wrote:

(In reply to comment #10)

Will it work properly in href-links too?

The semantics are the same, per the standards. I don't know if all user-agents will treat them the same. If any don't, that's probably a bug on their part.

Created attachment 5883
Patch adds rel=alternate and languages onto links.

This patch doesn't seem to trigger the appropriate mechanisms in SeaMonkey.

Attached:

ayg wrote:

+ title="<?php echo $langlink['text'];?>"

Doesn't this just repeat the text of the link? Why is it necessary? Also, doesn't it need to be htmlspecialchars()ed? (I don't think anyone's come up with a language called Bjegka"fka0###&@!! yet, but it's probably just a matter of time.)

+ lang="<?php echo $langlink['interwiki'];?>"
+ hreflang="<?php echo $langlink['interwiki'];?>"
+ xml:lang="<?php echo $langlink['interwiki'];?>"

Okay, that's just *excessive*. Only one of these should be needed. Or maybe two, if you really think it's important to mark up the fact that the word is in the same language as its target. We don't mark up different languages anywhere else in the interface, is there any good reason for it? It's going to add up to quite a lot of extra markup clutter on large articles . . .

(In reply to comment #13)

+ title="<?php echo $langlink['text'];?>"

Doesn't this just repeat the text of the link? Why is it necessary? Also,
doesn't it need to be htmlspecialchars()ed? (I don't think anyone's come up
with a language called Bjegka"fka0###&@!! yet, but it's probably just a matter
of time.)

Ahem. I am far too used to having abstraction libraries.

sumanah wrote:

Andrew, am I right in seeing that Aryeh did review your patch?

  • Bug 32725 has been marked as a duplicate of this bug. ***
Nemo_bis renamed this task from Add metainformation to pages with interlanguage links to Add metainformation for interlanguage links (link rel="alternate" hreflang).Feb 9 2015, 3:35 PM
Nemo_bis updated the task description. (Show Details)
Nemo_bis set Security to None.

I think this should be closed in favor of T93213.

I think this should be closed in favor of T93213.

Thanks for point out the existence of a pseudo-duplicate. That report is focused on a very specific implementation (<head>) which has specific issues. Brion's comment on the revert is still relevant, we should probably start by adding rel=alternate to the sidebar links.

Example current HTML from the English Wikipedia:

<li class="interlanguage-link interwiki-fr"><a href="//fr.wikipedia.org/wiki/Communaut%C3%A9_de_communes_de_la_Champagne_balgycienne" title="Communauté de communes de la Champagne balgycienne – French" hreflang="fr" lang="fr">Français</a></li>

See my comment at https://phabricator.wikimedia.org/T93213#4438137 re interlanguage links generated (or not) by Parsoid (or wikidata).

Aklapper changed the subtype of this task from "Task" to "Feature Request".Feb 4 2022, 11:02 AM