> -----Original Message-----
> From: Roland Bless [mailto:[email protected]]
> Sent: Thursday, September 29, 2011 1:15 AM
> To: Dan Wing
> Cc: 'Joel M. Halpern'; '6man'
> Subject: Re: Centrally assigned "ULAs" for automotives and other
> environments
> 
> Hi Dan,
> 
> On 28.09.2011 23:28, Dan Wing wrote:
> > ALGs are harmful and the NAT industry has over a decade experience
> > that shows ALGs are harmful.  ALGs have prevented proper operation
> > of SIP, FTP, and a variety of other protocols.  The more complex
> > a protocol, the more likely an ALG interferes with the complex
> > protocol -- rather than helping it.  This is because the ALG makes
> > naive assumptions of message flows and interfere with advanced
> > functions the protocol would like to do.
> 
> I know that very well, because I've attended your excellent tutorial
> at some IETF meeting in the past. :-)

Did I come across as disliking ALGs a bit?  ;-)

> > An ALG also requires unencrypted communications (so the application
> > can be examined) and, if the application payload is supposed to be
> > modified, also requires using no integrity checking.  That means
> > the entire system has a greater attack surface just to allow the
> > ALG to examine and to modify the packets in transit.
> > An ALG also complicates upgrading protocols.  Protocol changes have
> > to be done so they remain compatible with the remote system (always
> > a requirement) as well as with the ALG (which is a requirement
> because
> > of the ALG).  This increases the complexity to the protocol,
> especially
> > as the ALGs, themselves, evolve and have their own bugs fixed, but
> > are not proper, signaled elements in the architecture.
> 
> Agreed, but I think that the case of NAT-ALGs is a little bit different
> as they try to be transparent to the end-hosts.

Yes, that's the definition of an ALG.

There are ALGs for firewalls, as well.  The first ALG written, to my 
knowledge, was for a firewall (not a NAT) and was for "active mode" 
FTP.  Active mode FTP is where the server makes a TCP connection back
to the client.  If the client is behind a firewall (e.g., corporate
firewall), FTP broke.  Adding an ALG to the firewall, so it sniffed
FTP signaling and created a mapping in the firewall, allowed active
mode FTP to work.

> We think more of a
> security gateway/proxy architecture, where the existence of the proxy
> is explicitly modeled, e.g., use an HTTP proxy for web access.
> Sure, not all protocols allow the use of proxies.

But that's a proxy.  A proxy is not an ALG.

> Please note, that a car's onboard network is a very different
> use case than hosts operating in the open public Internet.
> We need a very secure solution in order to guarantee the safety
> of the car and the passengers. Which internal devices communicate
> externally is usually well-known in advance.
> Though a good point is that the protocols on the remote side may
> change and that you have to adapt to the changes.
> In some cases when using the proxy architecture, maybe only the proxy
> has to be upgraded,  not the internal devices, in other cases maybe
> the proxy in addition to the internal devices, which increases the
> complexity.

-d


--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to