I'm going to close this for now. Michael's theory is as close as I 
think we're going to get.

I remember that I once had a PHP/c-client combination that had a 
broken quoted-printable decoding function (imap_qprint).  Decoding a 
q-p file using imap_qprint() vs. a PHP based solution resulted in two 
different outputs.  Upgrading PHP and/or c-client fixed it for me.



In Horde 3.0, the QP decoding is done by PHP code so that could be why 
noone using 3.0 is seeing this problem (and explains why this works 
for some in 2.x but not others).

I don't think so, because I tested it with IMP 3 and it worked fine.

It is possible that this is something that IMP 3 simply doesn't handle 
well and that IMP 4 does. The MIME code is much improved and pretty 
much rewritten for Horde 3/IMP 4. I'd recommend trying it.

I have just realized what is happening.



The PDF document was attached with the following headers:



Content-Type: application/pdf

Content-Disposition: attachment; filename=DOCUMENT.PDF

Content-Transfer-Encoding: quoted-printable



and when it is downloaded, the quoted-printable encoding is not being 
converted back into a binary file.  If the document is saved from a 
mail client other than imp, it is converted back into a binary file 
from the quoted-printable format.  If the file is mailed with an 
encoding of base64, it is properly converted back to a binary file 
when it is downloaded from imp.



Have we configured something wrong and your test imp is doing the 
translation from quoted-printable back to a binary stream when the 
file is downloaded?

I'm not sure what else we can do on this one.

it works fine for me.

This question was to the other developers.

No, I'm dealing with an issue in our UVM footprints system that 
everyone who is receiving these messages from the Ariel system in our 
College of Medicine is unable to read them via our Horde/Imp 
installation.



So far the consistency of the report is that people using desktop imap 
clients are able to view the PDF file, but people who are using 
Horde/Imp (our installation at webmail.uvm.edu) are unable to view the 
PDF file because it gets translated when it is downloaded -- the OS 
and browser do not seem to matter.



Frank

Does this message work for everybody else?

Then I don't know what else to tell you.

I'm told we updated Horde to 2.2.7

Did you also update to the latest Horde 2 version?

Ouch.  We have patched our version of IMP to 3.2.7.  However, I'm 
still unable to view this file (in  the original message) using 
Firefox 1.0.  Is there something I need to check in my mime settings?   
Is there something that was changed that would not have been included 
in the patch deck to get from 3.2.5 to 3.2.7?

I can view and download this PDF fine with both the latest IMP3 and 
IMP4 versions and Firefox.

Oops... Sorry about that.  Sure, I can do that too.  Here it is.

I didn't mean the PDF file, I was talking about the *message source*.

Sure... Here's one of the files generated by Ariel.



The strange thing I found this morning is that saving the file (out of 
the original message) with Thunderbird and then creating a new message 
and attaching that file in Thunderbird, that the new message will 
properly display the PDF file from webmail...  I thought that was 
broken as well before, but I can't find any proof I tested it.



So, I'm not sure this how to test this anymore...

Can you upload the message source of such a message?

One of our major colleges has installed the Ariel program from 
Infotrieve.com to generate PDF documents.  The problem is that these 
PDF documents are damaged when they are viewed or downloaded from the 
IMP interface because they are being translated into a mime printable 
format (hex 0D is replaced by the ascii string =0D, etc).  PDF files 
generated by Openoffice are displayed just fine.



Using a standard mail client program (such as Thunderbird), one is 
able to view and save the PDF files without this translation happening 
to the Ariel generated files.



IMP shows the pdf.gif for both types of PDF files, but the download 
process seems to be very different.  This has been tested with Firefox 
on Fedora Core 3, MacOS X 10.3.7, Windows XP/SP2, also with Mozilla 
1.7 on RedHat 9 and IE on Windows XP.  And it always translates the 
Ariel generated PDF into a "damaged file" according to acrobat, xpdf, 
ggv, and MacOS X's preview.



Has anyone figured out how to provide support for the files Ariel is 
generating or is a newer version of IMP needed?



Thanks