Mailinglist Archive: opensuse (1480 mails)

[Peter Pöml <poeml@xxxxxxxxxxx>]

Hi,

Am 10.03.2010 um 13:27 schrieb Roger Oberholtzer:

> On Wed, 2010-03-10 at 12:39 +0100, Peter Poeml wrote:
> > aria2c.log

This confirms the suspicions I uttered earlier.

> I have attached the file. The download still failed. Popular wisdom (not
> final verdict) in the IT department is that nothing has changed. Still,
> I ran it on another computer on the same network, and it also failed. At
> home, with a different computer but same repos and, until this issue,
> the same software, it works fine.
>
> The odd thing is that it is failing so quickly. Not a tcp timeout. Some
> protocol failure.

That's because the protocol is manipulated by an intermediary that your IT 
department set up, as explained in the other mail.

They use a product called Ironport[1] that intercepts requests to the Internet, 
and interferes with them. It is like "forced" proxy, so to speak. Range 
requests are disabled in this proxy.

The IT department could change the configuration, by tweaking the 
"rangerequestdownload" parameter as described in the product documentation (I 
found one at 
http://www.headtechnology.ru/download/ironport/AsyncOS_Web_UserGuide.pdf - see 
p. 117). That configuration change would eliminate the issue.

Another possible configuration change might be that you bypass the proxy (by 
not configuring your client to use it), although I assume that's not possible 
for your client because the proxy intercepts the requests without the clients 
knowledge.

Short of changing the configuration, there are two things to examine.

1) Is Ironport's interception broken, in that it changes the passed headers to 
something invalid?
2) Is aria2c not correctly dealing with servers that ignore byte ranges

I guess that 1) is the case, because if Ironport removes byte range headers 
from the client, it clearly should also remove the server's announcement for 
support of byte ranges (which practically all mirrors announce): the 
"Accept-Ranges: bytes" header.

So IMO what we see is a violation of RFC2616 by the Ironport appliance. Might 
be worthwhile to open a bug with its manufacturer.

Peter

[1] http://www.ironport.com/products/web_security_appliances.html--
To unsubscribe, e-mail: opensuse+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: opensuse+help@xxxxxxxxxxxx