TIL - Podcast hosts serve unreliable art mime types

I’ve been toying with a small Django project that can be given a list of podcast feeds, and then regularly checks those feeds to update some interesting metrics. Initially, I wanted to see how many comparable podcasts to my own made use of various RSS namespace elements,¹ and also to pull some aggregate statistics around release frequency, episode length, etc.²

Since I was making a simple UI for myself, I decided to add in the ability to fetch and cache cover art of the podcast, updating it if the URL changes in subsequent checks. And here I noticed something odd. I was limiting the mime types that would be accepted in case garbage data was returned,³ but then I noticed that two podcasts in my initial set were never caching the artwork, even though the secondary checks on the file itself appeared showed valid types! When I dug deeper I found that the issue is with their hosts, which will happily return invalid or imprecise mime types in the Content-Type header.

For example, if the cover art uploaded to Spotify/Anchor ends in .jpg, it will deliver the file with mime type image/jpg, which does not exist. It should be sending it back as image/jpeg, and suggests to me they may be serving the designation by concatenating the file extension onto the the end of image/, which strikes me as a really bad idea. Captivate.fm, on the other hand, doesn’t even try to do the right thing, and reports any file type as application/octet-stream. 🤦

Now, I’m off to add some code to add a datapoint on these types of errors so that over time I can identify how many hosts are doing this!

UPDATE: Looks like Libsyn also does the application/octet-stream thing.