old versions of pages not rendering right
Status: Beta
Brought to you by:
worden
Menu
▾
▴
#150 old versions of pages not rendering right
Even without the full ability to reconstruct the old state of a project's files, I expect it to do a sort of cheap version - save the old text of the source files into the persistent directory and remake the output files. But it isn't working. When I go to a historical version of a page, I get the old wikitext with the current project file contents.
Discussion
Anonymous
yeah, looks like that feature dropped out during the refactor. I wonder if it'd be hard to put back in. It's kind of important for our reproducible research credibility - what's the point of saving old versions of papers when you can't see them? (There's a lively discussion in the open-science blog world right now about writing papers in the open and I'd like to point people to http://lalashan.mcmaster.ca/theobio/projects/index.php/Paper_for_MathWikis-2011 as an example, but I'm frustrated because the history part isn't working.)
This scares me. Without a complete reconstruction, I think I see two dangers.
1) A project may seem to be rendering the history correctly, but may in fact be showing the wrong version of some of the files, particularly if there are files that didn't used to exist, or are hosted on other pages, or whose location has moved ...
2) People may start squashing projects onto a single page for the purpose of taking advantage of cheap reconstruction.
who said I wasn't going to use old versions of files on other pages, and old versions of the project description?
thanks for pointing out the issue of files that didn't use to exist. it can be handled, but not if it's overlooked.
For complete reconstruction, see https://sourceforge.net/tracker/?func=detail&aid=3178719&group_id=366300&atid=1527388.
IIRC, you have asked me in fairly strong terms not to do this before finishing various other features. This puts me in a difficult in between place where my options are to do it partially or not at all, and not at all means we don't have the reproducibility features that are an important potential "selling" point.
(I notice your point (2) implies that it's a desirable feature that people will want to use)
Meanwhile, it can and should be implemented in stages.
(1) teach WW to locate the old content of source files and send it over to PE with the make request
(2) define semantics for specifying that the request is a historical one
(3) teach PE to handle historical requests in historical sessions.
Each of these can be done separately, producing a working feature at each stage (though it will work better after stage 3).
I also suggest stage 0: suppress misleading rendering for now. Xingpeng was really sandbagged by this one today. Because some project files were rendering and others weren't, he assumed incorrectly that reversion would not fix some issue he was having.
Was that clear? I mean the current behaviour, where the history page tries to render using current versions of files, is misleading in some situations.
Yes, this makes sense. It's a good idea - thanks.
We are now suppressing project files on old revisions of pages.
It mistakenly refuses to render on the diff page displaying the current version while comparing its text to an old version.
diff page problem fixed in r907