How do we deal with unknown mime types?

[tarekziade]

As seen on Fennec, some boxes just don't have mimetypes registered see:

https://github.com/pocmo/Fennec-DLC-Catalog/commit/b6abf51514f3d0a330166b2a4482a821cc86858c

we should deal with this in kinto-attachement. https://github.com/Kinto/kinto-attachment/blob/master/scripts/upload.py#L61

maybe by raising an error ? another option is to guess it on the server-side in Kinto when it's not provided by the client maybe ?

[Natim]

I think if no mimetype is provided in the client call we should probably fallback to application/octet-stream if we can guess with a streaming api it is fine but we if we need to open large file on the server side to guess, it can have a huge impact on performances.

That's why we wanted to have the client giving us the mimetype.

[tarekziade]

if we need to open large file on the server side to guess, it can have a huge impact on performances. That's why we wanted to have the client giving us the mimetype.

I am not sure that's relevant for the S3 backend (the performance part)

They don't enforce the content-type btw

http://docs.aws.amazon.com/AmazonS3/latest/API/RESTObjectPUT.html

I have not tested this but I think when you get a file in S3 where the Content-Type was not set, application/octet-stream is the default, so a fallback to that value seems a good idea

[leplatrem]

Does this only affects the upload script ?