Open mackenly opened 1 month ago
Hey @mackenly – thanks for submitting a PR!
I'm down with adding a POST
endpoint; makes total sense. I'm really hesitant to bifurcate the API though, where the GET
handler supports different parameter names than the POST
parameters.
I know the GET
parameters are pretty opaque (sid
, h
, p
, etc.), but I'd prefer we use the same names in both handlers. Similarly, what's the harm in also passing if-modified-since
as an HTTP header in the POST
request (vs passing it in the body
)? It should be possible to do so from the server-side.
Basically, I'd like to avoid complexity as much as possible, and this feels too much like having two very different handlers.
In some ways, it feels like the thing that this really needs is to accept and respond to a Content-type
(e.g. return the gif
or return a parseable JSON
response). Could we do that instead?
@benvinegar Good points! Would it make sense to have it at a separate path to clarify the difference?
Would it make sense to have it at a separate path to clarify the difference?
I think I'm trying to say: I'd prefer it didn't have a separate path, and instead the same original handler instead responds appropriately to Content-type
and returns a JSON response.
That means having to use the same hard-to-decipher parameters, and the same HTTP headers. That way we only have to test a single endpoint, document a single endpoint, etc.
Perhaps in the future I'd rename the parameters to be more human-readable, but we'd do it for every endpoint, not just POST
vs GET
.
Adds support for collecting events sent serverside. This enables users to send analytics events without running client-side scripts on their user's browsers. Furthermore, it will also enhance #17 allowing the managed component to use serverside tracking rather than the client side script.
Leaving this in draft for now. Still needs: