Question about change detection on older stories

I have the following feeds exposed on my blog:

  1. alexn.org/feeds/blog.xml
  2. alexn.org/feeds/all.xml

I’ve made several adjustments to them, and I noticed that there’s no way to update the items I’m getting in Newsblur. To make it clear — this is about refreshing older, already cached items.

The behaviour was strange, because feed #1 got refreshed last night only once, and then stopped refreshing, whereas feed #2 never did. I tried to force refresh it (via Insta-fetch, or by going into their Preferences and triggering “fetch feed”).

The other problem is that I’ve made small adjustments to older files, which ended up updating their dc:modified / atom:modified timestamp in the RSS feed — in feed #1, the one that was refreshed once, those feeds appeared as if they were new, published yesterday, with no link for viewing what changed. This got reflected in the feed’s sorted order too. They were detected somehow as new articles, even if their guid and their pubDate did not change. Whereas in feed #2 they were never detected as being changed, or updated for that matter, so the behaviour was inconsistent.

Questions:

  1. How can I invalidate Newsblur’s cache, such that I can ensure that other users get the right content (whenever I make adjustments to my feeds)?
  2. How does Newsblur detect new/unread articles?

Thanks,

Hi Alex, I renamed this thread to better capture the intent here. I think the answer to the question of how does NewsBlur detect changes in older stories, you can see story diffing checker here: NewsBlur/models.py at dashboard3 · samuelclay/NewsBlur · GitHub. Note that NewsBlur will only update 3 stories at a time, and sometimes more recent stories have a single character change, having to do with caching or ads or whatever, and that will trigger the update that then won’t propagate to older stories.

If you want to test this while I’m looking at it, go ahead and make changes to older stories and then go to beta.newsblur.com and insta-fetch stories from the feed. That way I’ll have a detailed log of what NewsBlur sees.

Also, any stories with code in them is not diffed, as to not mangle the code.