HttpClient 在 Content-Disposition 中编码错误

Question

I am POST-ing an image with HttpClient and it works well for files with Latin names, but as soon as a name contains any non-ASCII characters it gets transformed to a sequence of question marks. If I create an html form and use a browser to post the file, the file name is sent in UTF8 and the target server perfectly accepts it.

using (var client = new HttpClient())
{
    var streamContent = new StreamContent(someImageFileStream);
    streamContent.Headers.Add(
        "Content-Disposition",
        "form-data; name=\"image\"; filename=\"Тест.jpg\"");

    var content = new MultipartFormDataContent();
    content.Add(streamContent);

    await client.PostAsync("http://localhost.fiddler/", content);
}

This produces the following request:

POST http://localhost/ HTTP/1.1
Content-Type: multipart/form-data; boundary="e6fe89be-e652-4fe3-8859-8c7a339c5550"
Host: localhost
Content-Length: 10556

--e6fe89be-e652-4fe3-8859-8c7a339c5550
Content-Disposition: form-data; name="image"; filename="????.jpg"

...here goes the contents of the file...

I understand that HttpClient might work according to some standard, but anyway, is there any workaround?

UPDATE : The external API doesn't want to accept the format filename*=utf-8''Тест.jpg , it expects filename="Тест.jpg" .

Answer 1

OK, I've found a way to force MultipartFormDataContent to forget the ancient RFCs and use UTF8 instead. The trick is to use reflection to overwrite the DefaultHttpEncoding defined in the internal static class HttpRuleParser .

typeof(HttpClient)
  .Assembly
  .GetType("System.Net.Http.HttpRuleParser")
  .GetField("DefaultHttpEncoding", BindingFlags.Static | BindingFlags.NonPublic)
  .SetValue(null, System.Text.Encoding.UTF8);

Not sure which bad consequences that might cause, but I suppose there are none.

Answer 2

This is another way to workaround the limitation of HttpClient without tampering with internal fields. Inspired by this answer .

using (var client = new HttpClient())
{
    var streamContent = new StreamContent(someImageFileStream);
    streamContent.Headers.Add("Content-Disposition",
        new string(Encoding.UTF8.GetBytes("form-data; name=\"image\"; filename=\"Тест.jpg\"").
        Select(b => (char)b).ToArray()));

    var content = new MultipartFormDataContent();
    content.Add(streamContent);
    await client.PostAsync("http://localhost.fiddler/", content);
}

I confirm that even .net core 2.2 doesn't have proper support for uploading files whose names contain non-ASCII characters. HttpClient does work according to some standard but Java servers don't care about that standard and expect UTF-8 formatted headers.

Answer 3

Instead of adding a header that you built yourself, use the .NET library:

streamContent.Headers.ContentDisposition = 
    new System.Net.Http.Headers.ContentDispositionHeaderValue("form-data") { 
        Name = "image", 
        FileName = "Тест.jpg" };

That creates the header per the web docs and RFC5987 .

Content-Disposition: form-data; name=image; filename="=?utf-8?B?0KLQtdGB0YIuanBn?="

Answer 4

If it helps, you can also remove the " filename *"

//It deletes filename* parametr
foreach (var content in multipartContent) {
   var headerContent = content.Headers.ContentDisposition.Parameters.Where(x => x.Name == "filename*").SingleOrDefault();
   if(headerContent != null)
      content.Headers.ContentDisposition.Parameters.Remove(headerContent);
}