NoDogSplash (NDS) is a high performance, small footprint Captive Portal, offering by default a simple splash page restricted Internet connection. From version 5.0.0 the Forwarding Authentication Service API (FAS) has been moved to the openNDS package. This allows NoDogSplash to be optimised for use on devices with very limited resources (eg 8/64 or less). NoDogSplash v5.0.0 and greater will only support templated html splash pages.
For details of the OpenNDS package, see: https://openwrt.org/docs/guide-user/services/captive-portal/opennds
Full documentation can be found at:
The official repository can be found here:
NoDogSplash is a Captive Portal Engine. Any Captive Portal, including NDS, will have two main components:
A wireless router will typically be running OpenWrt or some other Linux distribution.
A router, by definition, will have two or more interfaces, at least one to connect to the wide area network (WAN) or Internet feed, and at least one connecting to the local area network (LAN).
Each LAN interface must also act as the Default IP Gateway for its LAN, ideally with the interface serving IP addresses to client devices using DHCP.
Multiple LAN interfaces can be combined into a single bridge interface. For example, ethernet, 2.4Ghz and 5Ghz networks are typically combined into a single bridge interface. Logical interface names will be assigned such as eth0, wlan0, wlan1 etc. with the combined bridge interface named as br-lan.
NDS will manage one or more of them of them. This will typically be br-lan, the bridge to both the wireless and wired LAN, but could be, for example, wlan0 if you wanted NDS to work just on the wireless interface.
By default, NDS blocks everything, but intercepts port 80 requests.
An initial port 80 request will be generated on a client device, usually automatically by the client device's built in Captive Portal Detection (CPD), or possibly by the user manually browsing to an http web page.
This request will of course be routed by the client device to the Default Gateway of the local network. The Default Gateway will, as we have seen, be the router interface that NDS is managing.
As soon as this initial port 80 request is received on the default gateway interface, NDS will “Capture” it, make a note of the client device identity, allocate a unique token for the client device, then redirect the client browser to the Portal component of NDS.
The client browser is redirected to the Portal component. This is a web service that is configured to know how to communicate with the core engine of NDS. This is commonly known as the Splash Page.
NDS has its own web server built in and this can be used to serve the Portal “Splash” pages to the client browser, or a separate web server can be used.
The default installation contains a fully functional splash page with a “click to continue” button. This is served from the included splash.html and splash.css files and can be easily customised by editing these files.
From version 5.0.0, the OpenNDS package is required for FAS support.
All modern mobile devices, most desktop operating systems and most browsers now have a CPD process that automatically issues a port 80 request on connection to a network. NDS detects this and serves a special “splash” web page to the connecting client device.
The port 80 html request made by the client CPD can be one of many vendor specific URLs.
Typical CPD URLs used are, for example:
It is important to remember that CPD is designed primarily for mobile devices to automatically detect the presence of a portal and to trigger the login page, without having to resort to breaking SSL/TLS security by requiring the portal to redirect port 443 for example.
Just about all current CPD implementations work very well but some compromises are necessary depending on the application.
The vast majority of devices attaching to a typical Captive Portal are mobile devices. CPD works well giving the initial login page.
For a typical guest wifi, eg a coffee shop, bar, club, hotel etc., a device connects, the Internet is accessed for a while, then the user takes the device out of range.
When taken out of range, a typical mobile device begins periodically polling the wireless spectrum for SSIDs that it knows about to try to obtain a connection again, subject to timeouts to preserve battery life.
Most Captive Portals have a session duration limit (NDS included).
If a previously logged in device returns to within the coverage of the portal, the previously used SSID is recognised and CPD is triggered and tests for an Internet connection in the normal way. Within the session duration limit of the portal, the Internet connection will be established, if the session has expired, the splash page will be displayed again.
Early mobile device implementations of CPD used to poll their detection URL at regular intervals, typically around 30 to 300 seconds. This would trigger the Portal splash page quite quickly if the device stayed in range and the session limit had been reached.
However it was very quickly realised that this polling kept the WiFi on the device enabled continuously having a very negative effect on battery life, so this polling whilst connected was either increased to a very long interval or removed all together (depending on vendor) to preserve battery charge. As most mobile devices come and go into and out of range, this is not an issue.
A common issue raised is:
*My devices show the splash page when they first connect, but when the authorization expires, they just announce there is no internet connection. I have to make them “forget” the wireless network to see the splash page again. Is this how is it supposed to work?*
The workaround is as described in the issue, or even just manually disconnecting or turning WiFi off and on will simulate a “going out of range”, initialising an immediate trigger of the CPD. One or any combination of these workarounds should work, again depending on the particular vendor's implementation of CPD.
In contrast, most laptop/desktop operating systems, and browser versions for these still implement CPD polling whilst online as battery considerations are not so important.
For example, Gnome desktop has its own built in CPD browser with a default interval of 300 seconds. Firefox also defaults to something like 300 seconds. Windows 10 is similar.
This IS how it is supposed to work, but does involve some compromises.
The best solution is to set the session timeout to a value greater than the expected length of time a client device is likely to be present. Experience shows a limit of 24 hours covers most situations eg bars, clubs, coffee shops, motels etc. If for example an hotel has guests regularly staying for a few days, then increase the session timeout as required.
Staff at the venue could have their devices added to the Trusted List if appropriate, but experience shows, it is better not to do this as they very soon learn what to do and can help guests who encounter the issue. (Anything that reduces support calls is good!)