Re: [WikimediaMobile] Separate tags for Apps and Web?

I strongly advise against using edit summaries. If we want to be able to filter and aggregate on revtags, parsing edit summaries is a very inefficient way to store this information; it’s also abusing the purpose of a field that was not meant to store revision metadata (the fact that antivandal tools do so is because of the lack of alternatives, not because it’s a best practice). If we want to collect granular data about the edit interface or the client used to complete an edit, the only options on the table are (1) MediaWiki tags or (2) instrumentation via EventLogging. MediaWiki tags • unless you concatenate in a single tag all relevant metadata (uggh) , they will only capture some very high level property of the revision • they are public data, exposed in revision histories, recentchanges etc • they provide hook for functionality (useful if we think people will build gadgets) • there are community expectations on what tags are good for or not that we need to be aware of EventLogging instrumentation • it would allow to capture granular information about the edit in a structured way and without the limitations of MediaWiki tags • data collected via EventLogging is currently not published/replicated to labs. This is not because the information collected is private based on the terms of the privacy policy but because we haven’t worked yet on a solution to expose this data. • EL data cannot be read off from MediaWiki, so if the plan is to enable functionality in MediaWiki based on tags, this is not an option Of course a combination of the two is also possible. The relevant questions to ask are: • how much information needs to be captured in a tag? • is this data meant to be public? • are tags expected to provide hooks for functionality in MediaWiki or are they for analytics purposes only? Dario PS the same problems of what tags to use applies to account registrations, not just revisions On May 14, 2014, at 9:41 AM, Maryana Pinchuk <mpinchuk(a)wikimedia.org> wrote: