I just checked locally and the large images downloaded file and had the Content-Disposition header.
Then I went to @Sam’s blog, clicked on the graph on this post and it also downloaded as a file. I checked the headers in Chrome and it too had the Content-Disposition
this is somehow related to the Rails Engine being mounted. Its was very hard to debug when I started looking adding it.
You should be able to reproduce this here somehow as well, since we are using akismet. Even if you bypass the cdn for the links
EDIT
@eviltrout header appears to be missing here https://meta.discourse.org//assets-meta-cdck-prod-meta.s3.dualstack.us-west-1.amazonaws.com/original/3X/d/9/d9988c49657be1d3ade114885f695a1aa25f71ef.gif
Just to say I don’t know what might “mount an engine” - but this occurs on my instance which doesn’t have “tagger”, nor “akismet” but does have the following:
It looks to me like the posts that can’t be downloaded on Sam’s site don’t have the data-download-href attribute. That attribute was added about a year after that post, so I think a rebake will fix it.
I can confirm that a post from “Sep 12, 2014” fails to download as expected and a post from today works as expected on my instance.
Rebaking the individual (previously failing) post fixes the download link for all images in the post.
Looks like it was added here (Oct 15, 2014):
So in-terms of rebaking (if you do confirm):
Clear instructions to rebake perhaps should be added for the point version upgrade it effects?
Rebake all button in the backend (with confirmation).
Handling future required rebakes - possible way of flagging upgrades / updates as requiring rebaking posts.
Rebake impact question - I’ve never done a rebake only the baking that was done automatically on forum import.
The baking process at the time of import actually took ~7 days - rebaking worries me as to it’s impact on the forum…
… it continuing to be usable and backups not causing an issue when running for multiple days.