[squid-users] No protocol or hostname part of url problem

I'm hoping someone can help me figure out a resolution to a problem we are
having. Preferably a Squid-based solution, although I'll take anything that
gets this working. :-)

We have 2 users who use a PDA-based application to perform safety audits
on-site. When they return to their office, they connect the PDA to their
computer, log in to the application provider's web-site with their browser,
then, using the PDA, upload the results of the audit to the provider.

We just deployed a Squid proxy out there this past weekend, and they cannot
now upload to the provider. In investigating the problem, I have discovered
that the PDA application does not properly form the POST urls. What the PDA
sends as the POST url is on the order of '/url/form.asp' instead of
'http://host.dom.ain/url/form.asp'.

In examining the flow of events on the wire, it looks like the reason it
works without Squid is that the HTTP packet has the host: header properly
set. It looks like Squid doesn't examine that header if there is no
hostname in the url.

Here's a tcpdump of one of the problem packets, with relevant parts
commented to protect the innocent:

POST /comm/login.asp HTTP/1.1
Content-Type: application/x-www-form-urlencoded; charset=unicode
Accept-Charset: unicode
User-Agent: CeHttp
Host: xxxxxxx.xxxxxxxxx.com
Content-Length: 34
Connection: Keep-Alive
Cache-Control: no-cache

sqsuser=xxxxxxx&sqspass=xxxxxxx

I know the most proper answer to solving this is to see if the application
provider has a fix for their broken application, but in the meantime, I
have these users constantly bothering me about not being able to send their
data. Anyone have any suggestions that could make Squid process these urls?

Thanks,
Will

-- 
----------------------------------------------------------------------------
William K. Hardeman
wont-i@wkh.org
http://www.wkh.org