Page MenuHomePhabricator

Track most-subscribed-to/most-tokened/high-priority tasks on Phabricator
Closed, DeclinedPublicFeature

Description

Feature summary (what you would like to be able to do and where):

Wikimedia should set up a dashboard to track the most-subscribed-to/most-tagged/high-priority tasks on Phabricator.

This would be done by making queries to Maniphest's tables since the Phabricator web interface has no way to search/sort through "tasks with the most subscribers".

Use case(s) (list the steps that you performed to discover that problem, and describe the actual underlying problem which you want to solve. Do not describe only a solution):

A tool for the WMF to evaluate popular Phabricator tasks and consider them when making development plans.

As @xavidegr commented in the proposal to "Dismantle the annual Wishlist survey system":

There are years of pending coding work to do with explicit, precise, extraordinary well-defined tickets both here (especially on the sister projects) and on Phabricator.

This thought, and my personal experience with popular tasks on Phabricator, is that there are tons of tasks Phabricator that have an extremely large following with a large number of subscribers, but they sit dormant every year and only are recognized when someone eventually comments on them, which may be a period of months or years. As a result, these are issues are not worked on and fail to be considered in the development plans of WMF teams every year.

WMF has the Community Wishlist Survey to evaluate community software needs, however that only covers features happened to be proposed for that CWS edition. There are plenty of proposals that may have high significance that have tasks here on Phabricator or were proposed in previous editions of the CWS, but happened to not be proposed for the current edition of the CWS and as a result, they may continue to remain dormant without consideration or action.

WMF needs to develop a system to track long-standing issues with software so that they can be evaluated against all other proposals in a holistic manner.
Creating a dashboard to track popular Phabricator tasks could be a first step in doing this.

Benefits (why should this be implemented?):

Better allocation of development resources towards highly-needed tasks that make large impacts.

Event Timeline

Lectrician1 renamed this task from Track most-subscribed-to/most-tagged/high-priority tasks on Phabricator to Track most-subscribed-to/most-tokened/high-priority tasks on Phabricator.Feb 15 2023, 3:08 PM
Lectrician1 created this task.

The most-tokened tasks are listed here: https://phabricator.wikimedia.org/token/leaders/ (it's not possible to filter by project; T302850 is for that).

mpopov subscribed.

Not sure why my team was tagged. I also want to note (as it seems historically relevant) that WMF actually developed a tool a few years ago called Phlogiston that generated reports from Phabricator data, but nobody used it and it was sunset (cf. T253964: Clarify status and future of Phlogiston (and sunset it if there are no users))

Peachey88 subscribed.

The Bitgeria instance that DA maintains may already cover this (Without pulling it up to have a look).

But worth noting Subscribers and/or token count does not equal importance in any way, and there can be many reasons why a task has high numbers.

would be done by making queries to Maniphest's tables since the Phabricator web interface has no way

I would like to point out that:

This was the outcome of a previous task when someone asked for access to the DB for analytics (T99295)

  • we already have a system that allows members of a special group live access to the database (without having full root on the production machines)

This automatically sets up a my.cnf for acces and all that and is used by @Aklapper . It's not hard to add people to this group if they are comfortable using shell.

# mysql read access for phab admins, in production (T238425)
if $::realm == 'production' {
    $::admin::data['groups']['phabricator-admin']['members'].each |String $user| {
        file { "/home/${user}/.my.cnf":.

Not sure why my team was tagged.

I am probably missing something but fwiw I would have also thought that https://www.mediawiki.org/wiki/Product_Analytics matches this kind of thing.

Phlogiston

Indeed, this used to be a thing but was shut down in T253964.

Bitergia

Does WMF still dedicate resources to keep this going?

Launchpad (a bug tracker used and created by Canonical/Ubuntu) has a concept of "bug heat" that is a similar indicator based on a few metrics.

But, I think it'd be worth reaching out to them to see how effective that measurement was in prioritizing work.

"Most subscribed to" is a terrible metric because it's easily gamed. Anyone (logged in) can involuntarily subscribe as many people as they want to a task. If you tell me that T89970 needs to have a hundred subscribers to win, I can accomplish that in less than five minutes, without consent (or often even knowledge) of the people I "subscribe" to it.

Aklapper closed this task as Declined.EditedFeb 25 2023, 12:04 PM

A tool for the WMF to evaluate popular Phabricator tasks and consider them when making development plans.

This task seems to propose a solution which I don't consider a solution to the underlying problem (if I understood the problem definition correctly), with some fallacies: Subscribing only implies interest at the moment that you manually subscribed && if you still follow your notifications nowadays && if your account is still active.
I don't think that thumbs-down tokens imply "popularity" - tokens have no clear well-defined meaning.
Priority is not consistently used && anyone can set Priority as part of wishful thinking. There are many ancient tasks with a "high priority" for >8 years.
In an open system like Phabricator anyone can game the system by creating accounts.
I personally do not believe that popularity contests on discrete tasks are a good base for planning software development as they may ignore the bigger picture. See also many debates in the archives of the wikitech-l@ mailing list about "voting" in Bugzilla (before 2014). And some posts out there about "design by committee" being a bad idea.
I don't see new arguments provided so I'm going to decline this. I encourage folks to propose tasks they consider popular themselves for the Developer Wishlist.