Skip to main content
Another small win today.

#Oni, the single user federated instance based on #GoActivityPub, serves HTML and #ActivityPub content from the same URLs based on what the clients request.

Firefox was caching the responses without taking into consideration the Content-Type, and was serving the JSON instead of HTML if refreshing a page. Only today I was reminded of the Vary header and fixed the problem. Yay! 😁

I'm a little bit ashamed of myself that I've forgotten about it.
#activitypub #Oni #GoActivityPub
Hypolite Petovan
Hypolite Petovan  
•
@m😛etagram:rius Coincidentally enough, we had the same issue recently in #Friendica : https://github.com/friendica/friendica/pull/13233
#friendica @marius
marius
marius  
•
old things are new again. 😁 Old knowledge resurrected.
Hypolite Petovan
Hypolite Petovan  
•
@m😛etagram:rius The issue only cropped up with some reverse proxies which none of the main devs use on their node 🤷‍♂️
@marius
marius
marius  
•
and are you guys sure it's the Accept header that needs to be taken into account for caching and not Content-Type ?

Vary is supposed to contain a list of the headers that the proxy/user-agent need to take into account when constructing the hash key for storing resources. Usually it's just the URL, vary adds a new element in the matrix for each header that it includes.

Accept is a request header, so I suspect that a vary with it will have no effect.

https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Vary

Vary - HTTP | MDN

The Vary HTTP response header describes the parts of the request message aside from the method and URL that influenced the content of the response it occurs in. Most often, this is used to create a cache key when content negotiation is in use.
developer.mozilla.org
marius
marius  
•
I am probably wrong and vary: accept works fine for proxies. My reasoning above might be correct only for user-agents...
Hypolite Petovan
Hypolite Petovan  
•
@m😛etagram:rius From what I know reverse proxies insert themselves between clients and servers. The Vary header is used by the proxy from the server to figure out what requests from the client will trigger a cache hit or a miss. In this context, it has to be a client header (here Accept) because the proxy cannot know whether there are multiple Content-Type values for the same URL from the server.
@marius
Hypolite Petovan
Hypolite Petovan  
•
@m😛etagram:rius Thinking more about it, I believe it’s the same for browsers, they need to know what request headers will produce different renditions of the same URL to be able to correctly cache it.
@marius
marius
marius  
•
yes, I've tested with my production environment and vary: accept works as expected, but vary: content-type doesn't.

I wonder if I gaslit myself, or I had different tests in my dev environment.
This entry was edited (1 year ago)
Hypolite Petovan
Hypolite Petovan  
•
@m😛etagram:rius Now I’m super glad I pointed out the coincidence in Friendica! 😄
@marius
marius
marius  
•
if I can make a suggestion, you should return raw images for Image type objects if you save them on Friendica.

I like this feature a lot in #oni:
https://marius.federated.id/outbox/1/object
Hypolite Petovan
Hypolite Petovan  
•
@m😛etagram:rius Thank you for the suggestion, I’ll pass it along to @Michael Vogel who is the most knowledgeable in protocol implementation.
@Michael Vogel @marius
⇧