Ticket #362 (Invalid)Thu Oct 31 15:35:50 UTC 2013
Are wiki pages really vanishing?
Reported by: | Trevor Johnson (329) | Severity: | Critical |
Part: | Web site: Instiki (documentation Wiki) | Release: | 3rd public site release |
Milestone: | 3rd public site release completed | Status | Invalid |
Details by Trevor Johnson (329):
- Some wiki pages, e.g. Bounty proposal: Paint and VFPSupport, have apparently vanished.
- Do ROOL have a backup they could be recovered from?
- They’re not listed in All Pages; do site admins have access to a deleted list, in case they’ve somehow been reclassified as deleted (if Instiki provides such a feature)?
- Should people be wary of making changes to the wiki until this is resolved?
- It’d be nice if content could be archived at the Internet Archive, but that looks as though it may still only be compatible with MediaWiki.
Changelog:
Modified by Andrew Hodgkinson (6) Mon, December 02 2013 - 14:15:39 GMT
- Severity changed from Normal to Critical
We do have backups, but they’re rotated out quite quickly and I’ve not seen your fault report as I was travelling at the time it was made and didn’t get around to checking up on tickets until today. I fear it might be too late – however – do you know when one or more of the apparently missing pages was either created, or at least, known to exist? That way at least I have some place to start looking at archives.
I don’t know off the top of my head of anything that should allow pages to be deleted, unless they were orphaned. If orphaned, occasional “delete orphaned pages” sweeps might have caused them to be wiped. That’s conceptually fine, because orphaned pages have no “inward route” in the Wiki other than bolt-from-blue external links. Perhaps, though, some people have intentionally made orphaned pages for just that sort of usage.
Perhaps that’s what happened in this case?
(I’ve increased the severity for this report at least until the cause is understood because unexpected page deletion is clearly a very serious bug).
Modified by Jeffrey Lee (213) Mon, December 09 2013 - 14:33:13 GMT
As far as I know the two pages were just orphaned pages not linked from anywhere. Although it did surprise me that the VFPSupport page had been around for several months before it went missing (although I guess it’s possible it had been missing for longer than I knew).
The VFPSupport page has now been constructed (and is no longer orphaned) so you shouldn’t have to worry about recovering that.
The Paint bounty page on the other hand hasn’t been recreated yet; a few bits and pieces have been collected together on the forums but I think there’s still some information missing. It looks like the page went missing somewhere between the 18th and the 25th of October (see the linked forum thread).
Modified by Jeffrey Lee (213) Thu, February 13 2014 - 13:28:07 GMT
However, it looks like some of the software information pages have now gone missing:
https://www.riscosopen.org/forum/forums/4/topic…
This is despite there being links to the missing pages from the bottom of here, with the intent that those links would keep the pages alive.
Modified by Jeffrey Lee (213) Thu, February 13 2014 - 13:38:05 GMT
And yet, if you google for some of the missing pages, it will return links which work, despite the list of all pages claiming that the page doesn’t exist:
https://www.riscosopen.org/wiki/documentation/s…
I think it’s fairly safe to say the wiki is broken.
Modified by Jeffrey Lee (213) Thu, February 13 2014 - 14:29:13 GMT
Poking around in the internet archive, it looks like these pages might have gone missing around October last year – i.e. the same time the other pages went missing.
Modified by Andrew Hodgkinson (6) Wed, February 19 2014 - 23:52:31 GMT
I ran an orphaned pages sweep some time last year but I can’t remember exactly when – stands to reason it was October. I hadn’t considered people relying on orphaned pages staying persistent with bolt-from-blue links outside the Wiki pointing to them. Oops. Anything genuinely orphaned at that point would’ve been deleted.
I’m pretty sure I remember checking our own internal CMS for references to Wiki pages and making sure they all had references, but 3rd parties might’ve been pointing into the Wiki too. If that’s the case, there are two things to do:
- Always create a reference to the page elsewhere in the Wiki, so it’s a complete single tree of self-contained, fully navigable documents
- Recover if necessary the old pages from e.g. Wayback Machine, as I don’t think ROOL backup archives go back that far
There seems to have been a Wiki bug which it seems maybe swept some /revisions/ of pages. For example, this page is listed as “wanted” in the All Pages list at the time of writing, yet it seems to exist:
https://www.riscosopen.org/wiki/documentation/s…
…however if you visit the page’s history, you’ll see that revisions 1 and 2 generate an “Internal error”, which in this case is a kind of early-version-of-Rails equivalent of 404 Not Found for the case where the page exists, but a corresponding database record doesn’t. That’s why the page is not appearing in a conventional sense, even though you can manually drag it out.
I need to look further into that as attempting to create a page with the same title as one which already exists is likely to have Hilarious Consequences. I’m not too bothered about the underlying orphan sweep bug as I just won’t run that again, but there’s the question about whether or not we now have creeping corruption in the live data set, or whether the Wiki will be robust in the face of missing revisions for some pages.
Modified by Sprow (202) Sat, March 21 2020 - 18:31:27 GMT
- Status changed from Open to Invalid
Conclusion: pages aren’t vanishing, it was just unfortunate timing of Andrew running a cleanup.
Ticket #479 covers the ‘hilarious’ consequences of creating pages which create circles in the universe.