Community Wishlist Survey 2017/Archive/Create a tool for merging global accounts
Create a tool for merging global accounts
Not technically feasible
- Problem: There is no way to merge global accounts now. During the SUL finalisation it happened often, that one contributor ended up with more global accounts. There can be other reasons as well, why one person has more than one accounts, and would like to merge them. As a bureaucrat and recently a global renamer I received several requests from many users in the last 10 years to merge their accounts, and I promised them that it will be possible soon (it is relative, but in some years), and I was waited for this feature as it was planned in the frame of the SUL finalization. Until end of last year when the developers announced that this feature doesn't work trustfully and won't be deployed.
- Who would benefit: Contributors, who has more (global) accounts and would like to merge them.
- Proposed solution: Improve Special:GlobalUserMerge or Special:UserMerge until it works :)
- More comments: Before the voting phase I would like to see if this request is technically possible or not. Please investigate if this is simply impossible for Wikimedia wikis or we need "only" to improve our software/hardware infrastructure to solve the problems.
- Phabricator tickets: T49918
- Proposer: Samat (talk) 12:39, 18 November 2017 (UTC)
Discussion edit
@Legoktm (WMF), Hoo man, Keegan (WMF), MarcoAurelio, and Nemo bis: I would appreciate your expert opinion on this topic. Samat (talk) 12:39, 18 November 2017 (UTC)
- We worked a lot on this but irreversibly broke accounts trying to merge them, see phab:T104686. For the tool, see further phab:T156584 and the tag usermerge on Phabricator. Best, —DerHexer (Talk) 13:23, 18 November 2017 (UTC)
- I think I know the earlier cases and the current situation, but this means only that it is currently not working and disabled. It is not clear for me, that the request is absolute not possible or just relative ( = hard to solve). Samat (talk) 14:04, 18 November 2017 (UTC)
- @Samat: You will not get an answer. It is a question of cost and benefit. What is the benefit? For some time now all new account are global. Only very old account would profit, and of those only that account that are still active, and of those only the ones that still want it. Let's say for the sake of the argument, that these are 200 people who benefit. What is the cost? It has been tried before, accounts have been irreversibly broken. Before it is tried again you need to test a new solution. You cannot test on production, so you will not break any more accounts. You need to create a replica for testing: 800+ wikipedias, with dozens of millions of users, with billions of edits. This test environment would need a hardware as big as that of production. That would cost millions, So instead you will run it on a smaller hardware. It will still run, but slow. The test of a single unification might take 2 days to complete. To test all 200 accounts may take 400 days. But if after 399 days of testing the system breaks, you have to correct the unification software and than start over. The correction might work, but in the end you may find another error, and you start again... Yes it is possible, No it will never be done. --𝔊 (Gradzeichen Diſk✉Talk) 23:57, 18 November 2017 (UTC)
- Thank you for your explanation. It is sad, but rational. Samat (talk) 00:30, 19 November 2017 (UTC)
- The tool exists, but there's a lot of legacy in the databases that prevent the tool from working on old accounts sadly, which are the ones that would really benefit from this. See phab:T49918#1933044 for details. That said, I'd love that now that we know the issues, try to work on fixing them if at all possible. —MarcoAurelio (talk) 20:30, 19 November 2017 (UTC)
- Thank you for your explanation. It is sad, but rational. Samat (talk) 00:30, 19 November 2017 (UTC)
- @Samat: You will not get an answer. It is a question of cost and benefit. What is the benefit? For some time now all new account are global. Only very old account would profit, and of those only that account that are still active, and of those only the ones that still want it. Let's say for the sake of the argument, that these are 200 people who benefit. What is the cost? It has been tried before, accounts have been irreversibly broken. Before it is tried again you need to test a new solution. You cannot test on production, so you will not break any more accounts. You need to create a replica for testing: 800+ wikipedias, with dozens of millions of users, with billions of edits. This test environment would need a hardware as big as that of production. That would cost millions, So instead you will run it on a smaller hardware. It will still run, but slow. The test of a single unification might take 2 days to complete. To test all 200 accounts may take 400 days. But if after 399 days of testing the system breaks, you have to correct the unification software and than start over. The correction might work, but in the end you may find another error, and you start again... Yes it is possible, No it will never be done. --𝔊 (Gradzeichen Diſk✉Talk) 23:57, 18 November 2017 (UTC)
- I think I know the earlier cases and the current situation, but this means only that it is currently not working and disabled. It is not clear for me, that the request is absolute not possible or just relative ( = hard to solve). Samat (talk) 14:04, 18 November 2017 (UTC)
Hi Samat, it looks like this proposal isn't really technically feasible. I'm going to archive this before the voting phase. Sorry for the sad answer; thanks for participating in the survey. -- DannyH (WMF) (talk) 17:38, 20 November 2017 (UTC)