Am 28.09.2015 um 16:43 schrieb Thomas Douillard:
Daniel Wrote:
(*) This follows the principle of "magic is
bad, let people edit". Allowing
inconsistencies means we can detect errors by finding such inconsistencies.
Automatically enforcing consistency may lead to errors propagating out of view
of the curation process. The QA process on wikis is centered around edits, so
every change should be an edit. Using a bot to fill in missing "reverse" links
follows this idea. The fact that you found an issue with the data because you
saw a bot do an edit is an example of this principle working nicely.
That might prove to become a worser nightmare than the magic one ... It's seems
like refusing any kind of automation because it might surprise people for the
sake of exhausting them to let them do a lot of manual work.
I'm not arguing against "any" kind of automation. I'm arguing against
"invisible" automation baked into the backend software. We(*) very much
encourage "visible" automation under community control like bots and other
(semi-)automatic import tools like WiDaR.
-- daniel
(*) I'm part of the wikidata developer team, not an active member of the
community. I'm primarily speaking for myself here, from my personal experience
as a wikipedia and common admin. I know from past discussions that "bots over
magic" is considered Best Practice among the dev team, and I believe it's also
the approach preferred by the Wikidata community, but I cannot speak for them.
I'm not sure what you are arguing against here.
Are you arguing against any tool that makes inferences combining multiple
pieces of data in Wikidata? Would you also argue against this if the inferred
information is flagged in some way?
peter