Hi folks,
as of yesterday, the CommonsMetadata extension is active on all wikis. As the name suggests, it's optimized for providing API access to metadata for files on Commons, but it will also work for project-local uploads (e.g. fair use metadata on en.wp) -- provided that the templates used for describing metadata emit the same machine-readable information.
The format for emitting such machine-readable metadata is documented here: https://commons.wikimedia.org/wiki/Commons:Machine-readable_data
The documentation page for the extension (which could use some cleanup & examples BTW) is at: https://www.mediawiki.org/wiki/Extension:CommonsMetadata
Help modifying local templates accordingly is appreciated.
All best, Erik
Erik Moeller, 23/11/2013 00:05:
The documentation page for the extension (which could use some cleanup & examples BTW) is at: https://www.mediawiki.org/wiki/Extension:CommonsMetadata
Help modifying local templates accordingly is appreciated.
Ahem, "accordingly" how? That page has zero instructions, it claims it's designed to work with stuff as it already is. Two sections are linked: * https://commons.wikimedia.org/wiki/Commons:Machine-readable_data#Machine_rea... (which actually talks of 2 formats, not 1) * https://commons.wikimedia.org/wiki/Commons:Machine-readable_data#Machine_rea... So I guess you're asking to add <td> id attributes and <span class="licensetpl_XXX"> markers to all the information and license templates?
By the way, this implies changing tens of thousands of templates across all wikis, often protected or very esoteric... It would be about time we recognise that local uploads are too big an effort we ask to most "small" local communities, even just for licensing policy compliance, and we start disabling them, centralising this burden on Commons. https://meta.wikimedia.org/wiki/Local_uploads_policy I hope Commons multilingualism is on the radar: having ignored it for 9 years is one of the reasons we now have an explosive technical debt hitting us every time we try to do even "simple" things.
Nemo
Hi Nemo,
the link to the Commons page was the first one in the email, and that's what the "accordingly" referred to.
So I guess you're asking to add <td> id attributes and <span class="licensetpl_XXX"> markers to all the information and license templates?
Yes, adding the same markers as in the Commons templates, e.g. a <span class="licensetpl_short"> to wrap the short name of the license, should have the same effect on a local upload being as it does on the Commons version of the file.
By the way, this implies changing tens of thousands of templates across all wikis, often protected or very esoteric...
I'd guess there would be some quick wins on the most commonly used templates. In cases where it fails, the media viewer will just add an explicit link to the file description page, so it's not the end of the world.
It would be about time we recognise that local uploads are too big an effort we ask to most "small" local communities, even just for licensing policy compliance, and we start disabling them, centralising this burden on Commons. https://meta.wikimedia.org/wiki/Local_uploads_policy
Perhaps so, but that's a bigger discussion for another place and another time.
I hope Commons multilingualism is on the radar: having ignored it for 9 years is one of the reasons we now have an explosive technical debt hitting us every time we try to do even "simple" things.
The hardest problems take the longest to solve, and in a context where for most of the 9 years, we've been technically very resource-constrained (we exceeded 10 full-time MediaWiki developers only in 2010), it's no surprise that implementing a multilingual tagging system and other fundamental improvements to multi-language support in a single wiki installation have not been done yet, nor ought it be cause for complaint or concern.
That is especially so given that throughout that history, many people (yours truly included) have already done a significant amount of thinking about the nature of some of the necessary technical improvements. Daniel Kinzler's WikiWord [1], back in 2008, demonstrated that a lot of the necessary information for building a multilingual tagging system exists in the form of interlanguage links, a fact that influenced the design of Wikidata and the initial focus on language links, since this would enable the quick evolution of a multilingual corpus of concepts.
Prior to Wikidata, OmegaWiki demonstrated that concepts annotated and edited in a wiki-style fashion could potentially be used to describe images and other media in a multilingual manner. The first proposal for this was written by GerardM in 2005, something that he, Kipcool and others continued to pursue in OmegaWiki for years to come (OW now supports Commons media). [2] [3]
The Wikimedia community, too, has made many improvements to language support in Commons itself, from the language selection features that predated ULS (and are still the only mechanism for anonymous user language selection) to localization of templates. The Language Engineering team contributed massive improvements to the Translate extension (active on Commons) and ULS which, while it still has its kinks, provides many important features that improve multilingual collaboration.
It is true that a lot of these mechanisms are hackish, but again, we're solving hard problems, which frankly haven't been solved well by others yet. Rather than complaining about technical debt, I think we ought to celebrate how much work has already been done.
These foundations help us see a clearer picture of how to actually implement something like a tagging system (although many important questions have yet to be answered). Then, still, we will not be satisfied until we have better multilingual discussion, or open source machine translation, or ... Iterative progress is to be expected.
The Wikidata team sees multilanguage support for Commons as a potential near-term frontier, and I think that's the right group of people to think about the problem. Daniel has written up an initial metadata proposal back in June, here: https://commons.wikimedia.org/wiki/Commons:Wikidata_for_media_info
There's still lots of work on the core of Wikidata before we can really expect them to take this one, but I for one am optimistic, and proud that so much good work has already been done.
Erik
[1] http://brightbyte.de/page/WikiWord [2] https://meta.wikimedia.org/wiki/Using_OmegaWiki_for_Commons [3] http://omegawiki.blogspot.com/search/label/Commons
wikitech-ambassadors@lists.wikimedia.org