Re: no-transform & working group last call for -p1 et al

On 28/09/2013 12:13 a.m., Eliot Lear wrote:
> The no-transform directive forever has said that proxies MUST NOT 
> touch payload.
>
> Situation:
>
> Suppose there is malware on a web site and a proxy resides between the 
> client and server.
>
> Questions:
>
>  1. Why would the malware distributor NOT want to issue the
>     no-transform directive?  After all, they don't want their malware
>     removed.
>  2. Why would a proxy honor the directive, knowing that there is malware?
>
>
> My point: I wonder if the MUST is a bit too strong or whether a caveat 
> should be added around this.  (Maybe there is such a caveat and I've 
> just missed it?)
>
> Eliot
>

I was introduced to the no-transform directive by its use on a hospital 
network.

The medical teams exchanged TIFF files and other very high resolution 
imagery, which simply *had* to be at that high resolution to identify 
and highlight the fine-grained pixels representing diseases or 
unidentified abnormalities. It surprised me how many off the shelf 
products even in such a sealed environment as a hospital take it upon 
themselves to "optimize" bandwidth by reducing such imagery into 
JPG/GIF/PNG or such.

I have since also seen it in use on satellite imagery and my own clients 
in the movie making industry send images and videos across the web using 
it for very similar reasons. Although for these clients it is not 
firewalls and desktop AV scanner proxies doing the optimizing, but 
mobile teleco and national level proxies.

There is not a requirement to deliver the infected payload anywhere - 
only a prohibition on delivering a modified copy. AV scanners have the 
option of replacing the entire HTTP reply with a 500 status response 
indicating the problem for manual intervention when no-transform 
prohibits fixing the issue silently.

I would say MUST is about the right level of severity on that one.

Amos

Received on Friday, 27 September 2013 13:59:06 UTC