From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search

Some things in LiWa3/COIBot to look at


  1. report ip x.x.x.x" ? Pushed and looking to be sitting in the queue, will see how they go. Thanks.
  2. Would it be possible to set "monitor ip x.x.x.x reason" so where IP is found in a generated new report that new report sits as an open report for review?
  3. Disallow more than one Reporter for a certain number.
  4. see whether it is feasible to make reports on last # days.
  5. is it feasible/worth it to report on filter hits?
  6. see if we can turn back on the 'report on username<->domainname' hits for link additions
  7. can we make functionality to turn on reporting on username<->pagename (userreport)
  8. diacritics in COIBot - googlé.xyz, also for usernames: User:COIBot/UserReports/猿香蕉 (due to diff), see also username mangling in User:COIBot/LinkReports/
  9. auto-split additions-section into chunks.
  10. repair the missing 'user' in some line reports
  11. check redirect detection:
  12. XLinkBot has a mistake in recognising IP talkpages correctly, it leaves a welcomeg on IPv4, welcomeanon on IPv6.
  13. align accesslists
  14. reconsider bringing lists to Wikipdia - userwhitelists and linkwhitelists, etc.
  15. last check on changed discussion before saving on COIBot reporting (edit conflicts).
  16. overflow on records which have hundreds of link additions and 1 coi addition do not make sense in terms of overflow. is an example.
  17. - 'top edit, link removed' ...
  18. recheck the problem with 'pagename=<namespace>:<pagename>' in some reports, maybe a COIBot hack to clear this up (IIRC I already did the way that LiWa3 is storing data (but check), but there are still old records).
    1. COIBot does not split namespace and pagename, see e.g.
  19. See if XLinkBot can use the delta(size) for all edits reverted, instead of 'the offending one'.
  20. parse out *. for reports of domains into userspace pages (the LinkSummary template results in a link-addition record ...)
    1. COIBot should not parse the reports in its own userspace.
  21. Categorisation (see Beetstra talk).
  22. Look if we can generate a reasonable error message when COIBot fails to save a page (or even, work around the error).
  23. Make sure everything is translatable in the reports through local settings.
  24. nowiki the added links on datadump - they sometimes appear as parameters in a linked form (see datadump).
  25. Error in 'nr search' -
  26. Remove adsense data for (per request)
    1. Done in principle (delete from linkwatcher_adsense where domain like 'org.archive.web%'; delete from linkwatcher_adsense where domain like 'org.archive.wayback%';), but maybe I want to consider to build-in domains for which we do not collect adsenses in the first place. But should archival services not just ignore the webbugs?


  1. check coding of unicode urls
    1. done
  2. diacritics bug in XLinkBot.
    1. done
  3. hunt.
    1. done
  4. write 'datadump' routine in Commander.
    1. done
  5. See if LinkReporter has access to the subject (instead of the Q-number) on WikiData, and can report that.
    1. No, is not there.
  6. try to get rid of those strange additions to reports due to http://http:// and '*.blah' and '.'
    1. done the 'double http' one, the other one is a different issue and less important.
  7. //<server IP> results in spam-blacklist blocks on many IPS
    1. no more live link on server IP, functionality is minimal anyway as it is also covered by the /tracked in the summary template provided for the IP.
  8. Get the regex for parsing added LinkSummary/UserSummary etc. from the on-wiki settings
  9. Give COIBot the capability to poke all templates from a random page.
  10. create {{user:COIBot/pokepage}}, which you can fill with a pagename on /Poke and then COIBOT will take the current version and poke all off the templates it can eat. Useful if you want to refresh.
  11. offload loading on wikis
    1. Wikidata to low-priority queue
    2. Wikidata and en.wikipedia to private DiffReader.
  12. plain diffs have changed resulting in COIBot not reading link additions. Parsing moved to '', solved diff content reading problem.
  13. fix mediawiki RC feed problems for COIBot (like I did for linkwatcher?)
    1. Wikidata and en.wikipedia to private DiffReader.
  14. authentication for LiWa3 now also through on-wiki settings list of cloaks.
  15. massively sped up LinkSaver in COIBot by also making sure that other links in the reports were not counted when they are whitelisted.
  16. Remove adsense data for (per request).
    1. Done (delete from linkwatcher_adsense where domain like 'org.archive.web%'; delete from linkwatcher_adsense where domain like 'org.archive.wayback%';)
  17. have a look at the diffs in en:Wikipedia:WikiProject Spam/LinkReports/, there seem to be urls assigned that are not belonging.
  18. Look at the parsing of WD (see Beetstra talk).