Weird problem getting HTTP/1.0 to work in my emacs browser

William M. Perry (wmperry@mango.ucs.indiana.edu)
Sun, 18 Jul 1993 15:41:06 -0500


Ok, lets see if anyone out there can figure this one out. I'm just
slightly confused this problem. :)

I have been working on my emacs www browser, trying to get HTTP/1.0 to
work. Now, as I understand it, a valid HTRQ would look something like
this:

GET /somefile.html HTRQ
From: wmperry@mango.ucs.indiana.edu
Accept-encoding: x-compress; x-zip
Accept: Accept: application/postscript
Accept: application/dvi
Accept: application/latex
Accept: application/tex
Accept: application/texinfo
Accept: application/x-rtf
[...] more headers here
User-Agent: Emacs-W3 .904b

^^ Blank line to end header, ala rfc822

Problem is, on some servers this will return a truncated page. I
checked, and my process-sentinel isn't jumping the gun, the
network-stream had actually closed, its just that part of the file
didn't come through. If I go to a shell window and telnet to the host
and send it the string GET /somefile.html HTRQ it responds immediately
with the whole file. (As its supposed to, since this one I know
doesn't grok HTTP/1.0 fully)

Sooooo - is it the extra lines that don't get actually sent to the
network stream that are causing the process to die early? (Typical
request is ~750 bytes) Is there some sort of maximum string size that
you can send via process-send-string? (Doubtful, since it works fine
on lots of servers that understand the new protocol, and others that
don't) <<< On the odd chance any elisp hackers are reading. :)

Also - the files are always truncated in the same spot. (I don't see
how it could be net lag either - this crops up going to NCSA from IU,
with an average round trip time (according to ping) of 29
milliseconds)

Could this be a problem with the servers? (Doubtful since it works
when it only gets sent the one line) The only server I've been able
to find this on is www.ncsa.uiuc.edu - /evl/Welcome.html always gets
truncated after "The animations are in a RGB file format and".
(Although I just retrieved it, and it worked, retrieve it again, and
it gets truncated after 'format and')

As I said, very weird. 1 person has reported that the HTTP server
here at IU shows similar flakiness with this new HTTP support, but I
haven't been able to make it die yet.

If I take the call to w3-create-mime-request out and replace it with
just a plain old "GET file" it works great.

Anybody out there that can shed some light on this?

Thanks in advance,
Bill Perry