Alan Sparks writes:
> My tests show that it often (can't be sure of always) occurs with a PDF
> file. BUT, it doesn't happen with NS Communicator and Acrobat Reader...
> only the combo of IE and Acroread.
>
> I've long suspected that it had something to do with IE sniffing the data
> stream to figure out the data type. There's an old bug against IE that
> mentioned that IE does sniff the data stream and use its own ideas in
> deference to the stated MIME types and content dispositions sometimes. I've
> always wondered if that ties into this.
I would agree to this. I have also seen that for the Autodesk
Whip! Viewer. You need to give MSIE using the PDF plugin a filename
ending with .pdf whatsoever. It seems not sufficient to supply the
content-type and leave the URL to something like /files?give_me_my_pdf
(what I was doing with Apache::File). MSIE (or the PDF plugin, I
don't know) then tries to find out the content type by itself using
the agent 'contype' and then issues a third request. All that seems
have nothing to do with Accept-Ranges and the like. It's really a
drag...
Dirk
