A new look at the interwiki link

For a recent essay based on this, specifically targeting interlanguage links on a single Project, see A newer look at the interlanguage link.

The interwiki link is intended to link articles in the different languages together. It is one of the more important ways to link the different projects together. Additional information can be found in the articles through the interwikis.

This is an essay from 2005, since updated with a specific proposal for language links and a working extension.


The problems

edit

New articles

edit
interwiki links between all languages
Current situation:
interwiki links
between all languages
=> proposed links of all langages to one central point
Proposed situation:
links of all languages
to one central point

As the wikipedia projects grow, more articles are written about the same subject in different languages. When one InterWiki link is added in one project it will eventually be added to the other wikipedias as well. This is most often done by the interwiki.py bot.

In the case where 20 languages have an article on the same subject, each article should link to 19 other languages, so there are 20*19=380 links to edit and maintain. In case of 100 languages, the number of interwiki links will amount to nearly 10,000.

Articles that are disambiguated

edit

When an article is disambiguated, or split into parts, the existing interwiki links needs to be changed on all projects. This is either done by hand or by the interwiki.py bot. It takes a human decision to correct this situation and the consequence is that these kind of errors can be persistent.

One changing article name, will lead to 19 edits on other languages in case of 20 languages, and 99 edits in case of 100 languages.

Articles that refer to a different level of information

edit

When an article like w:en:Vulcan (Star Trek) does not have a corresponding article in another wikipedia, there may be a need to refer to an article with a higher abstraction level like w:de:Völker im Star-Trek-Universum. The German wikipedia does not want separate articles for each and every Star Trek race therefore there is a need for "many to one" relationships. It is possible to show these relations on the English wikipedia, but it is not possible to show all these relations on the German wikipedia.

The solution

edit

Ellywa wrote about the problems on the wikitech-l. her suggestion is really good. There was some opposition and, these have been addressed in this proposal.

Technically what is needed

edit

Instead of linking every article on every wikipedia to every other wikipedia article with the same subject, a simple table should be required on a single location. This could be something like commons. Every article on each wikipedia needs to have only one link to this central list of links. If a name of an article changes, or an article is splitted, the interwiki links only have to be corrected at this central location. One change requires only one edit (in stead of 19 or 99 in the examples above). In the optimal situation, the articles in the local wikipedia's automatically get a list of links to other wikipedia's, which is retreived from the central database.

User:GerardM suggets to add the interwiki links in a new table, this table can be populated from the current interwiki links. Links to other projects are handled seperately.

The new table could look exactly the same as the current list of interwiki links on a local wikipedia. In addition, some (1 or 2) sentences could be added to give the exact meaning, or indicating it is a disambiguation page or a list of persons or objects.

As a note, this is already hand implemented on the Wikisource, as the project has many multilanguage pages coexisting together. See: Admin Interlangs, Policy and guideline interlang, and the original What is Wikisource interlangs. Of course, templates can't be included across projects, but the general spirit is there.
As of August 2005 Wikisource has been split into different language subdomains. The interwiki links are setup the same as for other projects, including the problem described on this page. HenkvD 15:23, 10 September 2005 (UTC)[reply]

New articles

edit

When an article is linked to another article, it will inherit the links of the central article. At the same time a server side bot will add the new link to the articles in the other languages.

Articles that are disambiguated

edit

When an article is disambiguated, the interwiki is changed on the article and the server side bot will make the changes on all the articles for the different languages.

Articles that refer to a different level of information

edit

When multiple articles refer to only one article (the Völker im Star-Trek-Universum example) they will not be shown on the German article. When the German article DOES add an article, it HAS to be on the appropriate level w:en:Category:Star Trek races is an appropriate link. The crux here is that the project where the multitude refers to decides what the appropriate article is to show.

What would it look like?

edit

As people are commenting that the solution proposed here is bureaucratic and difficult for inexperienced users, here a proposal how a page on the new wikimedia project "interwiki", with interwiki links could look like. The upper section could have a free format shortly explaining the word, in various languages if required, the lower section consists of the well known interwiki links, exactly the way they are used now. The example is for the article "dog".

The contents on an interwiki website, about the dog

English: A '''dog''' is an animal living at peoples houses
Nederlands: Een '''hond''' is een huisdier
Deutsch: Der '''Haushund''' (Canis lupus familiaris) ist ein Haus-, Heim- und Nutztier

[[ar:كلب]]
[[an:Gos]]
[[ast:Perru]]
[[bg:Домашно куче]]
[[bm:Wùlù]]
[[ca:Gos]]
[[cy:ci]]
[[da:Hund]]
[[de:Haushund]]
[[en:Dog]]
[[es:Perro]]
[[eo:Hundo]]
[[fr:Chien]]
[[gd:Cu]]
[[he:כלב]]
[[hu:Kutya]]
[[is:Hundur]]
[[nl:Hond]]
[[no:Tamhund]]
[[ru:Собака]]
[[ja:イヌ]]
[[pl:Pies domowy]]
[[pt:Cão]]
[[simple:Dog]]
[[sr:Пас]]
[[fi:Koira]]
[[sv:Hund]]

The interwiki link on the various language pages

This would be very simple and would be the same on all languages:

[[interwiki:dog]]

Of course, it is not required that the central article should have an English title, but it seems to me in most case the most practicle way, similar to the growing practice in Commons. Animals and plants could have the scientific name. People and places the name in the original language, etc.

The benefits

edit

There are many people doing great work on maintaining these interwiki links. To do this a lot of bandwith is wasted in trawling time and again over the database. I know that at least 10 people in the nl:wikipedia are working on this. This represents a huge amount of bandwith.

Which will be greatly reduced by implementation of the plan below

With a server side bot, the changes are almost immediately implemented and the resulting quality will be of a higher standard. (the German article still has Vulcan links ..)

a server side bot can do the same thing, using the same method for finding links. A bad implementation, however, will make it impossible for people to solve incorrect links with the help of a bot. If implemented in a bad way, we will need guru's to solve interwiki problems, or just accept the quality of the databases deteriorates in time.

The people who spend so much effort on running the interwiki bot will be freed to do more productive work. NB please let it be understood that at this moment in time it is very productive work.

Alternative solution

edit

The above proposal has been criticised by interwiki bot operators', some of whom have suggested their views are not taken into account by this proposal.

An alternative plan was proposed by Quistnix on the wikibots-l mailing list. The alternative suggestion involves merging warnfiles before splitting them, which is Quistnix believes will ensure maximum effectiveness of interwiki bot operation.

It will reduce the number of steps involved for a complete update dramatically. It reduces a second order function to a first order function, ensuring it will work in a growing wiki.

A step further

edit

Interwiki links between different languages for say Wikipedia are currently maintained seperately on every language. Some bots are in place trying to keep every language up-to-date (or only their own).

It would be better to have all interwiki links on one central place. The Wikipedia:en has currently this informal role, but it would be much better to have a new project, much like Wikimedia Commons. The name for it could be Wikimedia Links. This should not only work for Wikipedia but for all other Wikimedia projects, like Commons, Wiktionary, Wikispecies, Wikinews, Wikibooks, Wikisource, Wikiquote and Wikimedia.

doing interwikis, but not familiar with the Interwiki bot page? How can you talk about cooperation, if you are excluding yourself from it? -- Quistnix 18:06, 12 Jun 2005 (UTC)

How should it work:

  • all pages should have only one interwiki-kind-of-link to Wikimedia links, like w:en:Physics to links:Physics
  • on the Wikimedia Links system all interwiki links are listed, like
w:de:Physik, w:en:Physics, w:fr:Physique and other languages of Wikipedia Wikipedia
but also b:de:Regal Physik, b:en:Physics and other languages of Wikibooks Wikibooks
and wikt:de:Physik, wikt:en:Physics and other languages of Wiktionay Wiktionary
and commons:Physics, etc. of Commons Commons
  • The software should now work like this:
  1. look for the interwiki-kind-of-link to Wikimedia links
  2. retreive the actual interwiki links
    1. all interlanguage links for its own project (say wikipedia)
    2. all interproject links for the current' language (like en:Physics to en:wikibooks and en:wiktionay)
    3. all interproject links that are language independent (like commons, wikispecies, wikimedia)
  3. One additional link to Wikimedia links to be able to maintain the interwiki links.

Preferably these links should be displayed in sequence like below:

The benefits for this are the ease of maintainability of interwiki links between every language or between projects. Additionally it is no longer needed to add in-text links to other projects like via the en:Template:Commons. In general this might also result in better cooperation between projects.

(proposal of HenkvD 17:59, 12 Jun 2005 (UTC))

Further reading

edit

The article A new look at the interwiki link (2nd phase) explains a more technical database point of view on how to tackle this.

Opinions on this proposal

edit
This is much to difficult for an ordinary user to grasp. People will be stopped from doing interwiki because it is made to difficult.
What do you do with articles that do not have an equivalent on en:wikipedia ? But are linked between eachoter? There are 1000's of those?
Conflicts about which name it should get on wikimedia links etc.
I foresee chaos and making wikimedia projects yet even more of an unfriendly environment for starting editors. This is thinking of a difficult solution just so as to have a new technical tool to play with and sacrificing usability in the meantime. Upgrading the bots to perform better seems a lot better solution. Waerth 18:22, 12 Jun 2005 (UTC)
Yes it is more difficult. So is Commons, especially for non-english speaking users. But it is still very useful. As for articles that do no have an equivalent on en:wikipedia: these can be added just as easily, either by an english name or by any other, which in time may get renamed. HenkvD 18:35, 12 Jun 2005 (UTC)
And who is going to adminster that? This is what I meant by more bureaucratic. Waerth 18:47, 12 Jun 2005 (UTC)
We all will work together, like on commons. I don't see the need for bureaucracy. HenkvD 18:49, 12 Jun 2005 (UTC)
Yes and for the same reasons I am against commons. It is to difficult for a normal wikipedian to grasp it all. Which will mean it will be maintained by an elite few. So if there are problems, they will not be solvable by everybody as they are now. As most people do not really feel to go and look outside of the pedia to fix a problem. Waerth 18:58, 12 Jun 2005 (UTC)
Commons is at this moment difficult to work with. A ten year old child from France wants to find content when she looks for a "cheval". To improve on this I proposed Using Ultimate Wiktionary for Commons. This will hopefully help to improve the current situation. GerardM 19:47, 12 Jun 2005 (UTC)
The suggestion HenkvD makes will lead to an Anglocentric view spreading all throughout Wikipedia. I think this is a bad idea. If everything *has* to be linked via a wiki project in the English language (at least, this is how I read the plan), it will make good use of interwiki links only possible for people with knowledge of English. This is very unfair to people living in some parts of the world. -- Quistnix 16:41, 14 Jun 2005 (UTC)
The language of the lemmas don't have to be English, but most likely that will be the case. As the contents is multilinqual (the interwiki links into all other languages) I think this will be more multilingual as Commons. Alternatively a title may include several languages (like Main Page / Hauptseite / Accueil / メインページ). HenkvD 17:53, 14 Jun 2005 (UTC)
There are 2 aspects to the problem: 1) how to solve it technically and 2) how to solve it for the users. Both things should be solved in their own best way. In applications there are usually different tiers (layers). I believe that it is not that hard to create a technical solution that will create all other interwiki links at the moment that just one interwiki link is created. For the users it won't be any harder then it is now. See also my comments on A new look at the interwiki link (2nd phase), which proposes a relational table to enable this. Of couse practical problems remain, but those aren't any harder then the current practicle problems with articles in different languages which are not really congruent. Those are not technical problems but liguistic or philosophical problems. Just check the interwiki links for en:Fox (disambiguation) and try to decide if the portugese and polish interwiki's are really congruent. Taka 07:43, 30 Jun 2005 (UTC)

Support

edit
  • I support, because the presnt situation escalates quickly as the number of active wikis grows. But I would like to make stuff language independant. Would it be possible to create a template on each wiki that would create/update the link on interwiki: ? TeunSpaans 06:19, 12 December 2005 (UTC)[reply]
  • A common store for translations will be also a important step for the Semantic MediaWiki Project. Because it seem to be likely that there are only vew people that can handle semantics well, semantics should not be done in a redundant way at all. I've written also an article about some days before. -- MovGP0
  • SUPPORT. 68.148.165.213 07:19, 26 July 2006 (UTC)[reply]
  • Strongly support. This is a brilliant move. It is akin to the introduction of categories back in 2004. --LA2 21:20, 10 June 2008 (UTC)[reply]
  • Strongly support. This is not only a brilliant move, it is also much needed, as anyone who works in multiple languages knows. Now that SUL is live, this is probably the most important thing that needs to be done in support of Wikimedia' s multilingual goals. Better yet, it also serves cooperation between sister projects in the same language. Dovi 07:12, 11 June 2008 (UTC)[reply]
  • Strongly supportI Think this central goes help the Wikimedia projects.Tosão 23:33, 15 July 2008 (UTC)[reply]
  • Strongly support I am dreaming for such a feature since more than 4 years. It's an evidence. I don't undersatand why it is not yet done. Arno Lagrange  10:29, 6 October 2008 (UTC)[reply]

Oppose

edit
  • Some pages are avaliable for some limited wikipedias only. en:Solar term is an example. For zh:, ko:, and ja:, there are 24 seperated pages for each Jieqi/Jeolgi/Sekki. It is not meaningful for en: for separated pages. And there is no english names for Jieqi.
  • en:Wright brothers is another example. Some wikipedias maintain separated page for Orville Wright and Wilbur Wright. What is the solution for this case?
  • Is there a solution for the "rename war"?

-- ChongDae 08:25, 27 Jun 2005 (UTC)

The level of detail is different now as well. The example above just give some articles which on en: are in lower detail as on some other languages. In most of the cases it is the other way around for the other languages. I think this is quite normal and no reason for concern.
Is there a solution for the "rename war": I hope that with some naming rules this will be kept in reasonal bounds. For the proper interwiki linking the name is not realy relevant. It could even be a sequence number, but not my preference. Renaming war is unwanted though. HenkvD 11:36, 27 Jun 2005 (UTC)

That's not a matter of detail. For example, ko:음력 1월 1일(1st day of 1st month in lunar calendar) is no meaning in english. But in the east asia, it is a beginning of new year(en:Chinese New Year). Other days in the lunar calendar is also important for history. ko:음력 4월 10일(10th day in 4th month) is a birthday of King Sejong. I think that it will not happen to make a 21st day of 2nd month in Hebrew calendar in the Korean wikipedia. -- ChongDae 18:08, 27 Jun 2005 (UTC)

Not everthing has to be translated into English. Nobody said so. Mostlikely many articles will have English titles, but taxonomy might be in Latin, names of persons and places in the local language etc. HenkvD 18:45, 27 Jun 2005 (UTC)
Maybe a bit off-topic, but it is a bit disappointing if the information about such a special day would not be translated some time in the future into other languages. This is exactly why the interwiki links are so important - because peoples with different mother tongues can learn from each other and may understand each other a bit better. The title of such an article could be open. It could be a transliteration, or it could be in the original characters. Actually, it doesn't matter. Similarly to how we have now articles about Kimchi in several western languages (see en:Kimchi and its interwiki links). The title could be anything, even an arbitrary number, but that is not so handy. Maybe you can think of a solution for such cases? Ellywa 21:18, 27 Jun 2005 (UTC)
Opposed. The problem is not technical, but linguistic and cultural: there is no reason why the concepts should be identically structured in all languages and cultures, and especially no reason why the reference should be english: this is cultural imperialism, negation of diversity, and negation of the aims of wikiproject. If the issue is purely technical, why isen't Esperanto proposed as a pivot language, for instance? Michelet-密是力 04:21, 15 May 2007 (UTC)[reply]
CAuse there's a loooot of English speakers more than Esperanto. Osias 13:42, 27 May 2007 (UTC)[reply]
The opposing voices above have misunderstood the purpose of this proposed solution. Nobody has required all languages to follow the same structure. Nobody has required English to be the dominating language. The proposed system doesn't change how interwiki links can be used to connect the same topic between languages. It only changes how these links are maintained. Today, if 100 languages share a concept (e.g. an article about the city Moscow), we need to insert 99 interwiki links in all 100 languages. With this proposal, each of the 100 languages would only need to insert one interwiki link and the remaining 98 would follow automatically. This proposal saves a lot of work. That's all. How can you oppose that? --LA2 16:49, 10 July 2008 (UTC)[reply]