[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Captive-portals] A new draft / idea - draft-wkumari-capport-icmp-unreach

Hi Dave,

Yes, any middlebox can block/redirect traffic. But all clients must obtain an IP address before they send anything out.
I would say that both DHCP and IPv6 RA are in our scope. GTP is not, and
in fact GTP captive portals (yes, these things do exist!) use
alternative means to notify the user. My personal experience was that
they'll send you an SMS.
DHCP is commonly used to provide various types of information about the
local network, so what we're proposing to do is very common DHCP, and
does not mix layers at all. See

On 05/01/2015 03:22 PM, Dave Dolson wrote:
Any middlebox between user and server can block traffic and redirect http.
An IETF protocol should not mix layering or be limited to specific access technology.

IPv6 has ‎autoconfiguration for obtaining IP addresses.

3GPP has specific GTP-based mechanisms for obtaining IP addresses.

So DHCP isn't even always used to access the internet.

David Dolson
Senior Software Architect, Sandvine Inc.
   Original Message
From: Yaron Sheffer
Sent: Friday, May 1, 2015 7:43 AM
To: Dave Dolson; Warren Kumari
Cc: [email protected]
Subject: Re: [Captive-portals] A new draft / idea - draft-wkumari-capport-icmp-unreach

As a user, I have never seen a case where the captive portal is not on
the same link as the client device, and never seen a case where an IP is
obtained using a non-DHCP mechanism when in a CP setting. Can you give
some concrete examples?


On 05/01/2015 05:49 AM, Dave Dolson wrote:
I think it would be desirable to keep the captive portal mechanism independent of DHCP. The captive portal enforcement need not be on the same link as the user device. Furthermore, DHCP is not the only way to obtain an IP address.

David Dolson
Senior Software Architect, Sandvine Inc.
    Original Message
From: Warren Kumari
Sent: Thursday, April 30, 2015 8:02 PM
To: Yaron Sheffer
Cc: [email protected]
Subject: Re: [Captive-portals] A new draft / idea - draft-wkumari-capport-icmp-unreach

On Thu, Apr 30, 2015 at 10:52 AM, Yaron Sheffer <[email protected]> wrote:
Hi Warren.

The security consideration (an attacker can send any arbitrary URL, and will
redirect you at will) seems like a showstopper to me.
Yeah, we were a bit freaked about that too :-)

I have removed the URL / URI stuff - now this is simply annotates
Destination Unreachable to explain that the reason for the Dest
Unreachable is because of a captive portal.

Also, once you have the DHCP mechanism, you can have client-initiated
communication to a well-defined interface, and you don't need to deal with
arbitrary connections being rejected. After all, the client needs to get an
IP address from DHCP before it can initiate any such connection.

One reason that this is still useful is that eventually your captive
portal connection will expire and the CP will close. If you have
gotten CP information from DHCP, and then start getting these
Destination Unreachables you will know to connect to the captive
portal URL and pay again....
So, basically, get CP information from DHCP and use it. After 4hours
(or whatever), you start getting these new unrechables and know to
connect and pay again...

I think that the security implications are now the same as for
"regular" (extended) Destination Unreachable.


Thank you very much for your feedback,


On 04/29/2015 11:32 PM, Warren Kumari wrote:
Hi y'all.

So, this short document discusses another way for Captive Portals to
inform users that they are behind a CP. Basically, it uses an extended
ICMP Destination Unreachable message to let the host know that the
reason it cannot reach a destination is because it is behind a CP and
also includes a URL to reach the CP.

This idea is mainly David's, I'm largely the editor. We'd really like
some feedback on this idea and the implementation we've written up.
Obviously the document would need a bunch more work, but we'd like to
get some idea if folk think this is worth pursuing.



It is somewhat similar to the wkumari-dhc-capport document, but solves
a different issue, in a different way - I think that they are
complementary documents.

Any feedback gratefully accepted.

I don't think the execution is relevant when it was obviously a bad
idea in the first place.
This is like putting rabid weasels in your pants, and later expressing
regret at having chosen those particular rabid weasels and that pair
of pants.

Captive-portals mailing list
[email protected]