According to RFC 2046 and RFC 7578, the request header of multipart/form-data should not include the charset.

[dongfangtianyu]

Expected behavior

HTTP Header

Content-Type: multipart/form-data; boundary=2W1aSJ1TtJC_jRaGnbotI-RaHchFMAO

Actual behavior

HTTP Header

Content-Type: multipart/form-data; boundary=2W1aSJ1TtJC_jRaGnbotI-RaHchFMAO; charset=UTF-8

Steps to reproduce the problem

In JMeter 5.6.3, the request header Content-Type for multipart/form-data is required to include ; charset=.

On some web server implementations, including charset in the request header Content-Type of multipart/form-data can result in parsing errors of the boundary, leading to a failure in sending form content.

Example 1: https://github.com/spring-projects/spring-framework/issues/21599
Example 2: https://bz.apache.org/bugzilla/show_bug.cgi?id=61384

According to the latest RFC specifications, such implementations are incorrect:

In RFC 2046 [4.1.2] :

• For text and its subtypes (e.g., text/plain), the charset parameter should be passed in the Content-Type.
• i.e., The HTTP body is entirely composed of characters specified in the charset.

multipart/form-data does not belong to the text subtype, and the HTTP body may contain both text and binary data.

In RFC 7578 [5.1.2], rules for form encoding (form-charset) are defined:

1. If multipart/form-data specifies a charset, it should be located in the HTTP body rather than the HTTP header.
2. If charset is not specified for multipart/form-data, UTF-8 is used by default.

Therefore, HTTP headers like the following are non-compliant with the specification (and cause errors in some web server behaviors):

Content-Type: multipart/form-data; boundary=2W1aSJ1TtJC_jRaGnbotI-RaHchFMAO; charset=UTF-8

Interestingly, this HTTP header is also non-compliant with the specification (but doesn't cause errors as it lacks a boundary):

Content-type: application/json; charset=utf-8

I am not yet familiar with JMeter. If my idea is wrong, please remind me and close this issue.

Thank you.

JMeter Version

5.6.3

Java Version

17

OS Version

No response

[vlsi]

I've filed the issue for HttpClient: https://issues.apache.org/jira/browse/HTTPCLIENT-2325

[dongfangtianyu]

If HTTPClient removes new BasicNameValuePair("charset", charsetCopy.name()) ， or moves it before new BasicNameValuePair("boundary", boundaryCopy), it should be able to solve this problem.

However, I still don't understand the reason behind insisting on calling multipartEntityBuilder.setCharset in JMeter. Please enlighten me, and I would be grateful.

[dongfangtianyu]

I might understand now. Thanks again.