This check feels redundant at this point and I’m not sure it’s adding value to users but there’s no harm either.

After reading the documentation, it’s unclear to me why the previous error (raised for providing more than one archive or definition) raised ERROR_CONTENT and this one raises BAD_REQUEST.

Quoting the definition :

12.1.1. sword:ErrorContent

The supplied format is not the same as that identified in the Packaging header and/or that supported by the server

Associated HTTP Status: 415 (Unsupported Media Type) or 406 (Not Acceptable)

12.1.3. sword:ErrorBadRequest

Some parameters sent with the POST were not understood. The server MUST also return a status code of 400 Bad Request.

I feel ERROR_CONTENT is probably the right one in all these cases.

As you are now raising an exception if filehandler is None at the beginning, I guess this assert could be removed.

I am quite confused by your answer. I am not sure it matters regarding the actual error code, but as far as I can understand, the SWORDv2 spec mandates that there is only two parts in this case:

The client can create a resource by POSTing a multipart mime message to the Col-IRI with two parts: An Atom Entry containing the metadata for the deposit (known as the Entry Part), and the binary content as the second part (known as the Media Part), […]

Was there a time where swh-deposit behaved differently?

I'm also getting confused and it feels like we're getting trapped in a loop.

After reading the documentation, it’s unclear to me why the previous error (raised
for providing more than one archive or definition)

The client can create a resource by POSTing a multipart mime message to the Col-IRI
with two parts: An Atom Entry containing the metadata for the deposit (known as the
Entry Part), and the binary content as the second part (known as the Media Part), […]

Tentatively trying to break the cycle (maybe? ;p).

That quotes replies it better than i did then.

Are you asking for us to align the error to be the same in both conditional (in the
current diff, lines 578, 563, 591?). Why not, but If so, that's a subject for another
diff (and to me all those feel like "Bad request" anyway).

Was there a time where swh-deposit behaved differently?

It's possible that at an early stage, plenty of issues regarding those steps were not
hardened enough (hence the multitude of conditionals here to alleviate this). But the
gist of the behavior must have been the same (since we used the same spec your are
quoting afair).