mime: we don't decode quoted-printable

[bradfitz]

App Engine has a current production bug where long text values (non-file parts) of blob
uploads are sent from Google's MIME encoder to the user's HTTP handler, with the MIME
bodies re-encoded as quoted-printable.

Go's mime/multipart package does not decode quoted-printable.

Demo: http://bradgaetest.appspot.com/

The bytes sent to the handler are:

"--0016e68ee29c5d515f04cedf6733\r\nContent-Type: text/plain;
charset=ISO-8859-1\r\nContent-Disposition: form-data;
name=text\r\nContent-Transfer-Encoding: quoted-printable\r\n\r\nwords words words words
words words words words words words words words wor=\r\nds words words words words words
words words words words words words words =\r\nwords words words words words words words
words words words words words wor=\r\nds words words words words words words words words
words words words words =\r\nwords words words words words words words words
words\r\n--0016e68ee29c5d515f04cedf6733\r\nContent-Type: text/plain;
charset=ISO-8859-1\r\nContent-Disposition: form-data;
name=submit\r\n\r\nSubmit\r\n--0016e68ee29c5d515f04cedf6733--"

That was captured with:

        var buf bytes.Buffer
        r.Body = ioutil.NopCloser(io.TeeReader(r.Body, &buf))
    _, other, err := blobstore.ParseUpload(r)

I propose adding a non-exported quotedPrintableReader to the multipart package for now,
so this could be cherry-picked into production, but perhaps later we should move the
reader type to the mime (not multipart) package, and then just use that reader from the
multipart package (it already depends on the mime package, so it wouldn't be a new
dependency).

[bradfitz]

Comment 1:

Labels changed: added appengine.

[bradfitz]

Comment 2:

This issue was closed by revision d32d1e0.

Status changed to Fixed.

[wathiede]

Comment 3:

Is there a reason you wouldn't implement quoted-printable as a package under encoding
and let the user do the decoding, much like you would with base64 encoded data in a
multipart mime message? Alternatively do you foresee multipart.newPart growing the
ability to transparently decode more types?

[bradfitz]

Comment 4:

Yeah, as I mentioned in comment 0, we might want to move this to a general place later.
I don't know whether we should expand this to handle more encodings.  I felt like this
one was special enough, though.  It broke common cases, preventing things like
http://golang.org/pkg/net/http/#Request.FormValue from working.

[gopherbot]

Comment 5 by runningmaster:

When I tryed to get HTML from MHT file (MIME HTML) I lost (in HTML code) all line
endings and spaces in the text (all words are joined).
Concept:
    mht, err := os.Open("example.mht")
    panicIfError(err)
    msg, err := mail.ReadMessage(mht)
    panicIfError(err)
    _, prm, err := mime.ParseMediaType(msg.Header.Get("Content-Type"))
    panicIfError(err)
    bnd := prm["boundary"]
    rdr := multipart.NewReader(msg.Body, bnd)
    prt, err := rdr.NextPart()
    panicIfError(err)
    buf := &bytes.Buffer{}
    io.Copy(buf, prt)
    err = ioutil.WriteFile("x.html", buf.Bytes(), os.ModePerm)
    panicIfError(err)

[gopherbot]

Comment 6 by runningmaster:

My trouble with extract HTML part from MIME HTML file.
Please see for details:
http://play.golang.org/p/CPJFI_qMM4

[bradfitz]

Comment 7:

This bug is closed and marked Fixed.

[gopherbot]

Comment 8 by runningmaster:

OK, https://golang.org/issue/4771