[ python-Bugs-1097597 ] SimpleHTTPServer sends wrong Content-Length header

[SourceForge.net]

Bugs item #1097597, was opened at 2005-01-07 03:34
Message generated for change (Comment added) made by irmen
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1097597&group_id=5470

Category: Extension Modules
Group: None
Status: Open
Resolution: None
Priority: 5
Submitted By: David Schachter (razmatazz)
Assigned to: Nobody/Anonymous (nobody)
Summary: SimpleHTTPServer sends wrong Content-Length header

Initial Comment:
On Microsoft Windows, text files use \r\n for newline.
The SimpleHTTPServer class's "send_head()" method opens
files with "r" or "rb" mode depending on the MIME type.
Files opened in "r" mode will have \r\n -> \n
translation performed automatically, so the stream of
bytes sent to the client will be smaller than the size
of the file on disk.

Unfortunately, the send_head() method sets the
Content-Length header using the file size on disk,
without compensating for the \r\n -> \n translation.

I remedied this on my copy thusly:

      if mode == "r":
        content = f.read()
        contentLength = str(len(content))
        f.seek(0)
      else:
        contentLength = str(os.fstat(f.fileno())[6])

      self.send_header("Content-Length", contentLength)
 
This may not be as inefficient as it seems: the entire
file was going to be read in anyway for the newline
translation.

Hmmm. The code could be slightly simpler:

     if mode == "r":
        contentLength = len(f.read())
        f.seek(0)
      else:
        contentLength = os.fstat(f.fileno())[6]

      self.send_header("Content-Length",
str(contentLength))


The documentation for SimpleHTTPServer in Python 2.3.4
for Windows says:

   A 'Content-type:' with the guessed content type is
   output, and then a blank line, signifying end of 
   headers, and then the contents of the file. The file
   is always opened in binary mode.

Actually, after Content-type, the Content-Length header
is sent.

It would probably be nice if "Content-Length" was
"Content-length" or if "Content-type" was
"Content-Type", for consistency. The latter is probably
best, per RFC 2016.


By the way, clients weren't caching the files I sent. I
added another line after the Content-Length handling:

      self.send_header("Expires", "Fri, 31 Dec 2100
12:00:00 GMT")

This is egregiously wrong in the general case and just
fine in my case.

----------------------------------------------------------------------

Comment By: Irmen de Jong (irmen)
Date: 2005-01-10 01:49

Message:
Logged In: YES 
user_id=129426

http://sourceforge.net/support/tracker.php?aid=839496 is a
better url because it has been moved.
I've added a comment to my patch, because I'm now quite sure
it is good after all.

----------------------------------------------------------------------

Comment By: Johannes Gijsbers (jlgijsbers)
Date: 2005-01-09 18:18

Message:
Logged In: YES 
user_id=469548

http://python.org/sf/839496 has a patch for this, but the
submitter isn't sure whetther it's correct. Maybe one of you
could take a look at it?

----------------------------------------------------------------------

Comment By: Josiah Carlson (josiahcarlson)
Date: 2005-01-09 10:26

Message:
Logged In: YES 
user_id=341410

Would it be wrong to open all files with a mode of 'rb',
regardless of file type?

While I don't know MIME embeddings all that well, I do have
experience with email and that most codecs that use MIME
embeddings (like base 64, 85, 95, etc.) are \r, \n and \r\n
agnostic..

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1097597&group_id=5470