Re: [Duplicity-talk] S3 connection timeout

[Mitchell Garnaat]

I think the differences in the MD5 hash is simply different representations.  The Content-MD5 header creates a Base64 encoded version of the binary hash.  Most command line versions of md5 produce the hex representation by default.

Mitch

On 10/29/07, Peter Schuller <address@hidden> wrote:

> What are the chances that there is some length calculation that is
> off-by-a-constant or off-by-a-factor, causing a mismatch between the
> actual amount of data sent and the amount of data expected by the s3
> server? I would not be surprised if overrunning expectation would be
> handled by closing the connection.

So just to follow up on this, although I am more and more convinced
that it's not a client software problem:

I checked one instance of failure, and the Content-Length header
exactly matches that of the file.

The Content-MD5 does *not* match the plain md5sum of the file. Is
there some data tacked on at the end for example that would account
for the md5sum being different? In this particular case I do believe
the request failed prior to the entire file being sent though, so it
can hardly be that.

The data sent as part of the body, at least the initial few bytes,
exactly match that of the file.

Initial request (sensitive data stripped):

69296 python2.5 GIO   fd 8 wrote 397 bytes
       "PUT STRIPPED\r
        Host: s3.amazonaws.com\r
        Accept-Encoding: identity\r
        Content-Length: 268410864\r
        Content-MD5: PRh0gFTyAzKSiwTAQMzC3A==\r
        Expect: 100-Continue\r
        Date: Mon, 29 Oct 2007 20:48:42 GMT\r
        Content-Type: application/octet-stream\r
        Authorization: STRIPPED\r
        \r

md5 of the file:

MD5 (/tmp/tfile) = 3d18748054f20332928b04c040ccc2dc

Size of the file: 268410864 Oct 29 21:50 /tmp/tfile

--
/ Peter Schuller

PGP userID: 0xE9758B7D or 'Peter Schuller <address@hidden>'
Key retrieval: Send an E-Mail to address@hidden
E-Mail: address@hidden Web: http://www.scode.org


_______________________________________________
Duplicity-talk mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/duplicity-talk