Re: File Upload

Well, the first thing to check is request_size. This gave me some troubles
a while ago for users running Pointcast. The default config files says:

# TAG: request_size
# Maximum allowed request size in kilobytes. If people are using
# POST to upload files, then set this to the largest acceptable
# filesize plus a few extra kbytes.
#
#request_size 100

There is also another setting that might help if you need to bypass
your Squid for uploads:

# TAG: passthrough_proxy
# Specify the name of a 'cache_host' listed above, or a hostname
# and port number where all non-GET (i.e. POST, PUT) requests
# should be forwarded to.
#
# Usage: passthrough_proxy cache_host
# passthrough_proxy host:port
#
#passthrough_proxy

        -Bill

On Fri, 12 Sep 1997, Rodney Barnett wrote:

>
> I've just subscribed to this list, so please excuse me if I violate local
> protocol. Normally, I'd read messages for a while before posting anything,
> but I'm under pressure to get a problem solved and it might be Squid that's
> the source.
>
> My question is are people successfully do file uploads through Squid?
>
> I have a packet trace that appears to show Netscape 4.x (not the latest,
> but I can't recall the exact version) sending up a Word document correctly
> (about 32K) and another packet trace collected simultaneously that appears
> to show Squid sending the first few bytes of the document and then stopping
> abruptly. The server on the receiving end appears to wait 50 to 60 seconds
> and then generates an error.
>
> This particular case is Squid 1.1.10 on Linux 2.0.23. In general, this
> set up has been running successfully for many months.
>
> I also have a packet trace showing Squid 1.1.16's output to the same server
> with a different Word document. This time we upload over 250K and then
> suddenly the remaining 400K or so is skipped and a few of the trailing
> upload headers (for some of the other fields from the HTML form) appear
> along with the final MIME boundary. The server pretty quickly recognizes
> the end of the input this time and generates an error. I don't have a
> trace of the input to Squid in this case, so I can't be sure that Netscape
> didn't mangle it.
>
> I expect that there's not enough information here for anyone to give me any
> definitive answers, but does any of it ring any bells for anyone?
>
> Thanks.
>
> Rodney
> rlb@us.teltech.com
>
Received on Fri Sep 12 1997 - 18:36:26 MDT