On the infallibility of Wikipedia edit histories

Question

This is not really a question about this site (though it does have some impact), but I'm not sure where exactly to raise it and I'd like to pose it to this community, which can probably / hopefully help.

Recently I discovered that the Wikipedia data I used to produce an Astronomy Stack Exchange answer I'm rather proud of was, in fact, incorrect. The Wikipedia page has since been corrected (and I'll update that answer shortly), but I'm worried about what happened when I dug deeper for the original data to see where exactly I went wrong.

In particular, I went to the corresponding page in the edit history of Wikipedia, and the data shown there are simply plain wrong; specifically, the data table in the Details section is missing the exponent markers and is therefore unusable. This is made very clear by the corresponding version of that page on the Wayback Machine, which has the original incorrect data with exponent markers present. This seems to be more of a formatting issue, since there are exponents in the plaintext source, but the html shown in the edit history is still wrong.

I'm rather disturbed by this episode. Wikipedia editability is always a problem when linking to it, and that is resolved by the edit history, which is always there to resolve ambiguities about what exactly was being linked to. If the edit history is fallible, though, it's a red flag on the whole mechanism.

So, my questions for this community:

• To be honest, I just wanted to vent.
• Can others confirm that this is indeed what's displayed?
• Is there anyone more familiar with the finer points of Wikipedia formatting who can clarify exactly what went on there?
• Is there a specific place in Wikipedia where I can raise this issue? This feels more like a general thing than something for the article's talk page.

And, on our side:

• Is there anything we can do to adapt our practices given that this can happen?

Answer 1

This seems to be more of a formatting issue, since there are exponents in the plaintext source, but the html shown in the edit history is still wrong.

This is a formatting issue. Wikipedia does not store the HTML. It instead stores a page's wiki markup text. The problem is that the meaning of that markup language changes. Old pages are rendered to HTML using the current definition of the markup language rather than the definition at the time the page was written.

That's exactly what happened here. The October 2013 version of the referenced wikipedia page used (for example) 1.7{{Esp|6}} to represent what we would write as $1.7\times10^6$ as LaTeX code to be rendered by MathJax. (What? No LaTeX on meta.physics.SE?) The Esp wiki markup template no longer exists; it now redirects to the ESp template, which apparently has something to do with semi-protected pages.

Contrast that to the Wayback Machine (web.archive.org) works. It captures and stores a page's HTML. That doesn't mean that old captured pages will always display correctly. Some elements of HTML4 marked as deprecated were marked as obsolete in HTML5, and apparently even more stuff will become obsolete with HTML6.

Answer 2

It's indeed a template issue, as Hammen writes. The edit histories of both pages and templates are preserved, but the old pages are displayed using the current templates. Which makes some sense from an editor's perspective: you see what the page will look like when/if you restore that version.

Some wikimedia extensions can display old pages with the old templates: BackwardsTimeTravel, which implements it for history pages, and the Memento extension, supporting the Memento protocol which allows you to browse the web at a chosen date in the past (or at least that small part of the web that supports the protocol). Not available on wikipedia or other wikimedia sites however, the only website listed is w3.org.

To me the use of such a poorly designed template seems more problematic. A missing or wrong template should not result in a different value being shown, and there is already a template offering way more formatting options that doesn't have that problem, Val: {{val|3.14e4}} will display 3.14 x 10⁴ when the template can be found, and nothing or an error message otherwise, but it won't display only 3.14. "Val" has plenty of people maintaining it, "Esp" seems to have been the work of one person who wanted a tiny bit more space between the x and the 10. Fortunately it's not popular, when the redirect was re-targeted in Aug 2015, it was used 76 times on wikipedia, 48 of those were in the article mentioned.

Edit: I doubt that raising these issues at wikipedia (one of the village pumps, policy pages?, noticeboards?, Jimmy Wales's talk page) would have any impact, because:
- editors are already aware that old pages may not appear correctly
- they usually look at the wikitext rather than the rendered page
- all the content is in this case still available, so verification is still possible
- wikipedia doesn't consider itself a reliable source, so concerns about using it as source, especially using the old versions as source won't carry much weight
- link-rot is a big problem, many of the sources used in old pages won't work anymore, making the incorrect display of the content a rather minor issue in comparison
- wikipedia doesn't keep all history: old versions of deleted pages and templates can't be retrieved, so the problem will always exist. Making these available to editors is on the list of perennial proposals, always comes back and always rejected.
- a proposal to install the memento extension could get some support, but there is so much criticism about almost every technical development coming from the WMF that most would prefer that they first fix all the existing problems (visual editor etc...)
- Basically, I don't think it would be considered an issue, at least not more than it was already. Most editors have a list of improvements they want to see, things that help editors (better wikipedia search for example). While this could make it somewhat easier to view old pages, editors look mostly at diffs (differences between versions), which are shown in wikitext anyway.
- on second thought, I may have overstated the "risks" of such a badly designed template. Problems with templates are more typically caused by editors pushing changes in existing templates against concensus. This one is used on some 40 articles, all about astronomy it seems. Probably the persons who wrote it, so they will be alerted when anything happens to it.