LSL configuration files

The liblsl library can be configured with configuration files.

Configuration File Locations

There are four possible locations where such a config file will be found by the library (liblsl):

  • /etc/lsl_api/lsl_api.cfg (Unix); C:\etc\lsl_api\lsl_api.cfg (Windows). This is a global directory that is visible to all applications from all users on a given computer.
  • ~/lsl_api/lsl_api.cfg (Unix); %HOMEPATH%\lsl_api\lsl_api.cfg (Windows). This is a user-specific directory that is visible for all applications of that user. No administrator rights are usually necessary to edit it.
  • $PWD/lsl_api.cfg. This is a program-specific location: the file needs to be in the working directory of the program (for example right next to the binary). This allows to customize LSL behavior on a program-by-program basis.

New in version 1.13.

LSLAPICFG
  • the file specified in the environment variable LSLAPICFG This is a shell-specific location. On Linux / OS X you can set the env variable either before the program invocation (LSLAPICFG=/tmp/specialconf1.cfg ./LabRecorder) or as an export (export LSLAPICFG=/tmp/specialconf1.cfg) to apply it to all programs started afterwards from this shell session.

For the above settings, the more local settings files have the higher precedence, so a program-specific file overrides a user-specific file, which overrides the global file.

Configuration File Contents

The configuration file is formatted like a Windows .ini file.

The following section contains the default settings of LSL that can be pasted into a file as-is, although it is usually a good idea to specify only the subset of parameters that you would like to override (to allow for future corrections to the default settings).

[ports]
; This port is used by machines to advertise and request streams.
MulticastPort = 16571

; This is where the range of ports to serve data and service information begins (growing upwards according to the PortRange).
BasePort = 16572

; Ports from the BasePort to BasePort+Portrange-1 are assigned to both TCP data ports (on the even ports, if the BasePort is odd)
; and UDP service ports (on odd ports, if BasePort is even); since these ports are occupied in pairs, there can effectively be
; PortRange/2 stream outlets coexisting on a single machine. A new outlet will occupy a successively higher pair of ports when
; lower ones are occupied. The number of coexistant outlets can be increased by increasing this number. However, note that if
; multicast and broadcast or all UDP transmission are disabled on some router, the peers will need to "manually" scan this range,
; which can be slow on such a network. Also note that, to communicate with external parties, the port range needs to be open in the
; respective firewall configurations.
PortRange = 32

; How to treat IPv6: can be "disable" (then only v4 is used), or "allow" (then both are used side by side) or "force" (then only v6 is used).
IPv6 = allow

[multicast]
; The scope within which one's outlets and inlets are visible to each other. This can be machine (local to the machine),
; link (local to the subnet), site (local to the site as defined by local policy), organization (e.g., campus), or global.
; Always use only the smallest scope that works for your goals. This setting effectively merges the contents of
; MachineAdresses, LinkAddresses, SiteAddresses, OrganizationAddresses, and GlobalAddresses, and sets the packet
; TTL to one of the settings: 0, 1, 24, 32, or 255. If you share streams with remote collaborators, consider using the
; KnownPeers setting under [lab] (thus listing their machines directly, which is more likely to work than internet-scale
; multi-casting). Another possibility is to use the AddressesOverride and TTLOverride settings to avoid pulling in every
; site at intermediate scopes.
ResolveScope = site

; These are the default address pools for VisibilityScope. The following lists of addresses are merged according
; to the VisibilityScope setting to yield the set of addresses considered for communication.
; Note that making an uninformed/unfortunate address choice can interfere with your site's operations.
MachineAddresses = {FF31:113D:6FDD:2C17:A643:FFE2:1BD1:3CD2}
LinkAddresses = {255.255.255.255, 224.0.0.183, FF02:113D:6FDD:2C17:A643:FFE2:1BD1:3CD2}
SiteAddresses = {239.255.172.215, FF05:113D:6FDD:2C17:A643:FFE2:1BD1:3CD2}
OrganizationAddresses = {239.192.172.215, FF08:113D:6FDD:2C17:A643:FFE2:1BD1:3CD2}
GlobalAddresses = {}

; This allows you to override the addresses calculated by VisibilityScope. To communicate conveniently wth a remote party without negotiating
; the involved hostnames, you may choose a privately agreed-on multicast address of the appropriate scope here.
AddressesOverride = {}

; This setting allows you to override the packet time-to-live setting. If you intend to use multicast with a custom address to conveniently
; communicate with a specific remote party, you may set this to a sufficiently high level (255 for international collaboration).
TTLOverride = -1

[lab]
; This setting mainly serves as a fallback in case that your network configuration does not permit multicast/broadcast communciation.
; By listing the names or IP addresses of your lab's machines here (both stream providers and stream users) and make the file available
; on all involved machines, you can bypass the need for multicasting. This setting can also be used to link a small collection of machines
; across the internet, provided that the firewall settings of each party permit communication (forward the BasePort to BasePort+PortRange ports).
KnownPeers = {}

; This is the default "vanilla" session id; modify it to logically isolate your recording acitities from others within the scope.
; The session id should not be relied on as a "password" to hide one's data from unpriviledged users; use operating-system and
; network settings for this purpose. Note that you machine still gets to see some traffic from other activities if within the scope.
SessionID = default

[log]
; the log level. Only messages at this level or below will be logged
; -2: errors
; -1: warnings
; 0: information
; 1-9: increasingly less important details
; When liblsl is built with LSL_DEBUGLOG=On, the levels 1-9 will log more details
; The log looks like this:
; Date and time          |  runtime  |  thread ID        | source location        | level | Log message
; 2020-11-11 08:54:08.696 (   3.132s) [        A5247740]   inlet_connection.cpp:46    WARN| The stream named 'timesync' can't be recovered automatically if its provider crashes because it doesn't have a unique source ID
level = -2
; liblsl can also write all log messages to a file specified here. By default, log messages will be appended
file =

Changing the port ranges

To change just the port range to, say 3051 - 3068, create a config file with the following content:

[ports]
MulticastPort = 3051
BasePort = 3052
PortRange = 16

This type of change would only be necessary if you can move LSL to a port range that is allowed through or forwarded by the router or firewall (or the administrator).

Changing the multicast scope

Under some circumstances your recording environment might include a large number of routers. Service discovery between routers is a case that is not handled particularly well by current network installations (it requires correct company-wide multicast settings), but in cases where it works, you can expand or contract the scope within which two machines will see each other’s streams.

The boundaries of these scopes are defined by the network administrators, but they have the common names machine, link, site, organization, and global.

The default scope used by LSL is site. To change it to organization, use a config file like the following one:

[multicast]
ResolveScope = organization

In some cases it can also be helpful to reduce the scope to link (which is the local router), for example when you have many concurrent recording operations that you would like to generally separate from each other (some one experimenter should not see the others’ streams). In a local lab, the lab.KnownPeers option is usually a better choice, though.

Usually this is not necessary because between-router multicast is often not configured properly anyway.

Note that under the hood the multicast scopes are implemented by sets of multicast addresses (which have the scope encoded in their address). Independently of the scope you can customize the addresses themselves, for example to adhere to local administrative rules. See the full config file for the relevant variable names.

Defining the Local Laboratory

It is possible to define what constitutes the local laboratory network in a very fine-grained manner, if necessary (for example if one router was shared between 10 labs, each of which involves a number of machines, or if a single recording operation is coordinated across the internet between countries).

There are two mechanisms for this.

The KnownPeers setting allows to explicitly list the IP addresses or hostnames of the involved machines.

The following file contains an example:

[lab]
KnownPeers = {192.168.1.17, 137.243.177.26, testing.ucsd.edu}

With this setting any type of service discovery issues due to router configuration can be worked around. Note that at the same time you might want to disable the multicast discovery by restricting the ResolveScope to machine (the local machine) if the goal is to prevent interference.

The other mechanism does not involve the physical machines but is a purely logical partitioning of the network into separate and independent recording environments. This is accomplished by assigning a non-default value to the SessionID option. You only ever see streams hosted by clients that have the same SessionID.

Below is an example:

[lab]
SessionID = lab-001b

This way, you can assign a different session id per machine, or per user, or per application to bypass any sort of unwanted stream visibility between concurrent recording operations. Note, that the SessionID is not a security feature, however.

You are still be able to intercept packets involved in a session that is not yours.

Logging

In case something’s not working and you want to investigate it or produce a helpful bug report, liblsl can log details of the normal operation and errors to the terminal or a log file:

[log]
level = 0
file = lsllog.txt