Last modified: 2014-11-20 23:42:32 UTC
Some files edited with Picasa apparently have "Picasa" as their author in the EXIF metadata. Example: https://www.mediawiki.org/wiki/File%3ADarkvector_screenshot.jpg Obviously, the author isn't Picasa, so I'm wondering if we should have some sort of blacklist in CommonsMetadata to ignore the author field if it matches things that we know not to be authors. It's probably better not to show anything rather than showing something we know to be untrue. (This is similar to bug 58195 except in this case there isn't any other authorship information in the wikitext.)
Other examples: https://www.mediawiki.org/wiki/File%3ADisappearing_Username-1.jpg "Picasa 2.7" as author https://www.mediawiki.org/wiki/File%3AExtreme-testing-language-engineering.svg "Created with Raphaël 2.1.0" as title https://www.mediawiki.org/wiki/File%3AFor_talk_simple_security.JPG "Picasa 2.6" as author https://www.mediawiki.org/wiki/File%3AMediaWiki_Homepage_Proposal.svg Short title Untitled Image title Generated with SwordSoft Layout https://www.mediawiki.org/wiki/File%3ARegular_expression_complexity_exploit.svg Short title Qt Svg Document Image title Generated with Qt
And more: https://wikimediafoundation.org/wiki/File%3AGilt_silver_jar_with_pattern_of_dancing_horses.jpg has "OLYMPUS DIGITAL CAMERA" as image title
I wonder if EXIF shouldn't be ignored completely. Mostly it seems to be autogenerated and less than helpful. E.g. some cameras apparently put something like IMG1234 to the title.
I personally don't have enough data to decide if it makes sense to ignore completely, but I trust your judgment on that. It does seem like we have a lot of false positives.
After encountering more and more of these, like https://de.wikibooks.org/wiki/Datei%3ABenutzerMKabel.jpg , I'm starting to agree with you. I still believe a handful of users (including me) curate their EXIF metadata, for example by adding information using a digital collection management software, and we should support that, but this is already done (or should be) at the time of upload by extracting that data and prefilling the fields. Using them afterwards through CommonsMetadata seems to be more trouble that it's worth. Adjusting the title of this request accordingly.