Save and close the file, then run boinc-client to have the server connect to [email protected].
sudo service boinc-client start
Step 4: Checking the New Host
Step 5: Checking the Server Status
To view the status of the server or tasks, use the command:
This command provides general information [email protected], and also reports current tasks running on that particular host.
The output of this command looks something like this:
boinccmd --get_simple_gui_info
======== Projects ========
1) -----------
name: [email protected]
master URL: http://setiathome.berkeley.edu/
username: stmiller
team_name: SETI.USA
resource share: 100.000000
user_total_credit: 33159.675770
user_expavg_credit: 1409.252845
host_total_credit: 0.000000
host_expavg_credit: 0.000000
nrpc_failures: 0
master_fetch_failures: 0
master fetch pending: no
scheduler RPC pending: no
trickle upload pending: no
attached via Account Manager: no
ended: no
suspended via GUI: no
don't request more work: no
disk usage: 0.000000
last RPC: 1423684749.199424
project files downloaded: 0.000000
GUI URL:
name: Message boards
description: Correspond with other users on the [email protected] message boards
URL: http://setiathome.berkeley.edu/forum_index.php
GUI URL:
name: Help
description: Ask questions and report problems
URL: http://setiathome.berkeley.edu/forum_help_desk.php
GUI URL:
name: Your account
description: View your account information
URL: http://setiathome.berkeley.edu/home.php
GUI URL:
name: Your preferences
description: View and modify your computing preferences
URL: http://setiathome.berkeley.edu/prefs.php?subset=global
GUI URL:
name: Your results
description: View your last week (or more) of computational results and work
URL: http://setiathome.berkeley.edu/results.php?userid=9351194
GUI URL:
name: Your computers
description: View a listing of all the computers on which you are running [email protected]
URL: http://setiathome.berkeley.edu/hosts_user.php?userid=9351194
GUI URL:
name: Your team
description: View information about your team: SETI.USA
URL: http://setiathome.berkeley.edu/team_display.php?teamid=115396
GUI URL:
name: Donate
description: Donate to [email protected]
URL: http://setiathome.berkeley.edu/sah_donate.php
======== Tasks ========
1) -----------
name: 25fe12ab.24545.17667.438086664204.12.226_0
WU name: 25fe12ab.24545.17667.438086664204.12.226
project URL: http://setiathome.berkeley.edu/
report deadline: Wed Mar 4 02:06:18 2015
ready to report: no
got server ack: no
final CPU time: 0.000000
state: downloaded
scheduler state: scheduled
exit_status: 0
signal: 0
suspended via GUI: no
active_task_state: EXECUTING
app version num: 701
checkpoint CPU time: 352.733700
current CPU time: 378.866400
fraction done: 0.088431
swap size: 110309376.000000
working set size: 40030207.999996
estimated CPU time remaining: 2505.901220
Installation.
The installation process for BOINC Manager version 6.4.1 is described below.
(UPD: some screenshots from version 6.6.28 have been replaced - where there have been changes compared to the old version).
We start the installer, we see the first configuration window:
The first path, program directory, is where the BOINC manager itself is installed. We leave by default.
The second way, data directory - where project data will be stored: calculation modules, tasks, etc. I recommend changing the default path so that everything related to BOINC is in the same folder.
The next three checkboxes are left to taste, but I remove the first one, "use screensaver". What would not interfere.
If during installation it writes Error reading setup initialization file, then we read
Short translation:
1. run the installer with the /a option
2. indicate on which disk windows is located
3. install as usual
That's it, the initial setup process is completed, and after the installation is completed, the program will ask you to restart your computer.
After the reboot, you will see something like this screen:
Therefore, we go to "Advanced" - "program settings" and start hosting there.
(advanced - options - general - language selection - this is if suddenly you have not decided on a Russian machine).
Settings:
General tab
* Choice of interface language
* Interval between manager attempts to connect to the project server. By default, the manager automatically connects to the server every 60 minutes: to send ready-made tasks and to receive new calculations. If you want - change, I left it by default.
* Run manager on system boot? I think it's logical to start. Although if you do not need a tray icon, then you can not run it.
* Enable exit message? If there is a daw, then if you want to exit the manager, the following window will pop up:
In this window, you can make the following settings:
* "Stop the calculation process and exit the manager." A handy question that appeared only in the 6th version of the BOINC manager.
Do not check the box - the manager only closes, the project continues to work.
Check the box - the project stops and the manager closes.
* "Remember my choice and do not ask this question again" - I think it's clear what the second checkbox affects.
Connection Settings Tab
"HTTP proxy" tab
SOCKS proxy tab
Selecting an item in the menu "Advanced" - "customer settings"
Processor Tab
* "When powered by batteries"- relevant for laptops and PCs with configured software to work with UPS. Most likely we are removing it, why do we need to put a battery on the laptop?
* "when the PC is used for work"- a very necessary jackdaw, we must put it. Otherwise BOINC will wait until you leave the computer for "3 minutes" (default).
* "Use GPU when computer is used for work"- Use a video card for calculations when you work? If yes, then the video card will perform calculations constantly, if not, it will wait until you leave the computer for "3 minutes" (by default).
* The following is a group of settings for aesthetes: you can set the schedule according to which BOINC will make calculations.
Even lower are "other options".
* "Switch between projects every 60 minutes". If you uploaded a batch of tasks from different projects for calculations (for example [email protected] and Primegrid), then the client will change projects every 60 minutes: an hour is considered [email protected], then the hour is considered Primegrid, and so on.
* On multiprocessor systems, use 100% of the processors.
If you have a quad-core processor, then 4 cores will be used, if you put 50% - 2 cores.
* use no more than 100% of the processor- by what percentage will BOINC evenly load the available cores.
Those. if you put 100% of the available cores and 100% of the processor, then all 4 cores of our quad will work at 100%.
And if you put 50% of the cores, 80% load - then in our quad two cores will be loaded at 80%.
Network settings tab
Common parameters. This is cache management.
We set (rather, untouchable :)) the speed of downloading and sending.
A very important point! Choose how many days to upload tasks.
I, having a permanent connection to the Internet, set the main buffer for 1 day, the additional buffer for 1 day.
If you have the Internet periodically - then put more days. Just pay attention to the deadline for sending, so that the tasks are not "overdue".
How does the manager know how many jobs to download?
At the beginning of work, and then at intervals of 4-5 days, the manager launches a processor benchmark. And based on the results obtained, it decides how many pieces of VU (work unit = Work Unit = VU) to load into the cache.
* do not check uploaded files- I don’t know myself, I’ll find the answer - I’ll update this paragraph in the article.
Connection settings.
These two checkboxes are relevant for periodic Internet connections. The manager will ask you before trying to connect to the server, and will also be able to disconnect himself after downloading new and sending already counted tasks.
Access to the Internet.
Here you can set the schedule by which the manager will go to the network (if you need it).
Do you need constant internet access?
No! The manager needs access to the network only for sending/receiving tasks. So theoretically, a computer can stand without the Internet as long as there is something to consider in the cache.
Disk and Memory tab
Here I am not an adviser: what numbers you need to put is up to you. I left everything as default. That's just the "period of saving to disk" I would increase from 60 to 300 seconds. Why pull the HDD once again?
Everything, the initial configuration of the manager is completed. Let's move on to connecting the project.
Connecting to a project
Select "service" - "add project".
Next, a list of projects appears. It is incomplete, on the contrary - only the most "travelling" ones are concentrated in it. the most popular and sane BOINC projects. If the project that you need is in the list - good, if not - just insert the desired project address in the bottom field and click next.
How can I find the address of the project? It is in the "header" of the topic on our forum. Just remember that not all projects that we write about on the forum are BOINC projects.
Message:
Message from server: No work sent
The server is not currently issuing jobs. It's okay, there are projects in which there are no tasks for half a year. Connect to another project for now.
Message:
You used the wrong URL for this project
The correct URL is http://www.cosmologyathome.org/
This means that since the moment you connected to the project, the server address has changed. Now you need to reconnect to the project. Those. delete the project and reconnect with the correct server address. Everything about what was written above referred to the project manager installed on one local machine. To "local manager".
If you have one car, then yes, it is easier for you to manage it. But if you have ten of them, then it's easier for you to connect each and manage them centrally using "BAM" - "Boink Account Manager". More about him.
Fully dedicated to BOINC issues.
Working in the IT department, I constantly observe computers that are idle for various organizational reasons and are very bored. The golden days of bitcoin mining on the CPU are over, and in search of a new useful business, I came to voluntary distributed computing, in particular, to the World Community Grid. First of all, the search for a cure for cancer was puzzled by a server from a cold standby and a low-priority virtual machine on a virtualization cluster. It is more difficult with workstations, they constantly come and go, on each one it takes a long time to install, configure, and then remove BOINC.
It was decided to assemble a live distribution kit with BOINC "embedded" and distribute it over the network. Turn on the computer, press F12, select the desired item - and you are already benefiting humanity!
Debian was chosen as the platform, which a) has been familiar with for a long time and b) has a wonderful manual on the right topic. Nevertheless, it was not without a rake, and in this case, almost every new rake means a rather lengthy reassembly of the image. I hope this post will save some admin time, and at the same time remind you of the existence of such a wonderful project as WCG.
Note that everything was done in a very closed environment, and it took very little attention to security. Perhaps, in your case, you will need to work on security more.
Preparation
The system consists of this:- Network Boot Server.
- NFS server
- assembly station
1. Network boot server. Everything was already ready for me, I had the configured TFTP and DHCP from the thin client project. If you don't have one, picking up a new one is easy. In a nutshell, install and run tftpd-hpa, and specify parameters 66 and 67 in DHCP. Just don't let anyone boot over the network (in my case, these are cadets), it can be dangerous. In addition to the BIOS, you can password protect part of the boot menu of the TFTP server.
2. NFS server. First, BOINC must be able to save its data while it is running. It is assumed that the local hard disk cannot be touched, therefore, via NFS, we will allow writing to a directory, for example, /srv/boinc-nfs . Here, each computer will create a subdirectory with a name that matches its MAC address. Secondly, the /srv/debian-live directory will contain the root FS for network boot. So:
mkdir /srv/debian-live mkdir /srv/boinc-nfs chown nobody:nogroup /srv/boinc-nfs chmod 755 /srv/boinc-nfs
In /etc/exports add:
/srv/boinc-nfs *(rw,sync,no_root_squash,no_subtree_check) /srv/debian-live *(ro,async,no_root_squash,no_subtree_check)
after which we restart the service (for some reason, the recommended exportfs -rv did not give me any result):
/etc/init.d/nfs-kernel-server restart
3. Assembly station. It's just a virtual machine with regular Debian Wheezy. The live-build package is installed and will do the bulk of the work. There must be internet.
Assembly process
We go to the assembly station.mkdir /srv/live-default && cd /srv/live-default
We create a basic config for our distribution by specifying the address of the NFS server:
lb config -b netboot --net-root-path "/srv/debian-live" --net-root-server "192.168.15.20"
Some directory tree is formed, placing different content in them, you can customize your build. We will add the following:
1. config/package-lists/boinc.list - a list of packages that will be needed in our build. We write to it:
boinc-client nfs-common
2. config/includes.chroot/etc/init.d/boinc-preps - an init script that will mount NFS, configure BOINC and change hostname detached state). In this script, you need to insert the address of your NFS and the addresses of the hosts from which passwordless management will be allowed. Script content:
#!/bin/bash ### BEGIN INIT INFO # Provides: boinc-preps # Required-Start: nfs-common # Required-Stop: # Should-Start: # Default-Start: 2 3 4 5 # Default-Stop: 0 1 6 # Short-Description: Various stuff for BOINC # Description: Various stuff for BOINC ### END INIT INFO PATH=/sbin:/usr/sbin:/bin:/usr/bin . /lib/init/vars.sh do_start () ( MYMAC=`ifconfig eth0 | grep -o -E "([[:xdigit:]](1,2):)(5)[[:xdigit:]]( 1,2)" | sed s/://g` ancien=`hostname` nouveau=DYNWCG-$MYMAC mkdir -p /mnt/boinc-nfs mount 192.168.15.20:/srv/boinc-nfs /mnt/boinc- nfs && mkdir -p /mnt/boinc-nfs/$MYMAC service boinc stop sed -i "s/^BOINC_DIR=.*/BOINC_DIR=\/mnt\/boinc-nfs\/$MYMAC/;s/^BOINC_USER= .*/BOINC_USER=\"root\"/" /etc/default/boinc-client echo "192.168.10.60" > /mnt/boinc-nfs/$MYMAC/remote_hosts.cfg echo "192.168.10.61" >> /mnt /boinc-nfs/$MYMAC/remote_hosts.cfg echo "" >> /mnt/boinc-nfs/$MYMAC/gui_rpc_auth.cfg for file in \ /etc/hostname \ /etc/hosts # you can add #/etc/ here ssh/ssh_host_rsa_key.pub \ #/etc/ssh/ssh_host_dsa_key.pub \ # if you need SSH do [ -f $file ] && sed -i.old -e "s:$ancien:$nouveau:g" $file done invoke -rc.d hostname.sh start invoke-rc.d networking force-reload service boinc start ) case "$1" in start|"") do_start ;; restart|reload|force-reload|status) echo "Error: argument "$1" not supported" >&2 exit 3 ;; stop) # NOP exit 3 ;; *) echo "Usage: ... " >&2 exit 3 ;; esac:
3. config/hooks/boinc-preps-init.chroot - a one-command script that will be executed during build and add boinc-preps from the previous paragraph to autoload:
#!/bin/sh update-rc.d boinc-preps defaults
After adding the necessary settings, we launch the assembly itself:
lb build
The resulting binary.netboot.tar file is of value to us. Unpack it to /srv:
cd /srv && tar -xvf live-default/binary.netboot.tar
It will unpack to /srv/debian-live (netboot filesystem root) and /srv/tftpboot (TFTP server files). In my case, the build station and the NFS server are the same machine, so /srv/debian-live is already in place.
The content of tftpboot is a ready-made boot menu, it needs to be placed on a TFTP server. I did not check its work, since I have a working TFTP server with its own menu, and from here I needed only part of the data. First, I copied all the files from tftpboot/live/ to the TFTP server to images/debian-live/ (relative to the root directory of the TFTP server). Secondly, from tftpboot/live.cfg I borrowed the text that adds a new menu item, while changing it to this state (you also need to specify the address of your NFS server here):
label live-686-pae menu label BOINC-live (686-pae) linux images/debian-live/vmlinuz1 initrd images/debian-live/initrd1.img append boot=live config nosplash root=/dev/nfs nfsroot=192.168. 15.20:/srv/debian-live
Now everything is in its place.
If after assembly you want to check the contents of the root FS, then it is not necessary to boot with it, you can simply mount it as a loop device:
mount -o loop,ro /srv/debian-live/live/filesystem.squashfs /mnt/squash/
If you need to rebuild the distribution with new options, you can either do lb clean --binary or lb clean before doing so.
Usage
- We turn on the computer and select boot over the network (usually just press F12 to do this).
- Depending on the boot menu, either select the “BOINC-live” item, or simply wait until it loads by timeout.
- In the command line that appears (if everything went as it should), we write sudo ifconfig (you do not need to enter a password) and write down the IP address.
- On the management computer (one of those that we specified in config/includes.chroot/etc/init.d/boinc-preps) run boinc-manager, click "Advanced - Change computer" (this button is only in "Full View" ). In this case, BOINC should not request any passwords from the user.
- After connecting, a wizard will appear in which you will need to select a project (in my case it is the World Community Grid) and enter your login/password.
This procedure only needs to be performed once for each computer (more precisely, for each MAC address). Even after returning to you after a long operation in another place, the computer will find its data on the NFS server by MAC address and continue working immediately after being turned on (only some tasks will already be overdue, but this is a trifle, it will receive new ones).
Results
What remains unresolved:- Autologin in WCG. Probably, you can sew the project config into the image so that you don’t enter anything with your hands at all, but in a reasonable time the solution didn’t google.
- Mail notifications. In order not to enter "sudo ifconfig", you can automatically send the address of the running computer to the admin. However, I did not do this, because it's crutches, and it's better to implement the previous point instead.
- The IP of the NFS server is listed twice, perhaps one of them can be removed.
- The MAC address of the eth0 interface is always used to create a computer directory. Whether that's good or not, I can't say for sure.
Don't let your computers get bored! And write comments, I will gladly answer or supplement the article.
They use the BOINC platform, which was developed about ten years ago at the University of California at Berkeley. The first step to participate in one of them is to install the BOINC Manager software.
Installing and configuring the BOINC client
After downloading and installing the client program, you can specify one or more projects that you want to join. Starting with the sixth version, the BOINC client supports flexible calculation settings that allow you to specify exactly what part of the hardware resources a person is willing to sacrifice for the needs of science. In addition, you can set the BOINC Manager to automatically pause when you start some programs, or only calculate during certain hours. Calculations can be completely or selectively suspended manually at any time and resumed just as easily.
CPU and GPU Compute Settings in the BOINC Client
In addition to the options of the client application itself, settings are provided for each project through the web interface on the participant's personal page.
BOINC project settings via web interfaceYou can evaluate the computing power of a computer with the built-in BOINC benchmark. It is also sometimes used to compare the actual performance of different configurations.
Embedded BOINC Client BenchmarkSelection of projects based on available resources
When performing scientific calculations, the performance of floating-point FPU units plays an important role. The accuracy of the calculations may be different. Usually they talk about three types implemented in practice:
- FP16 (Half Precision) - half precision of floating point calculations;
- FP32 (Single Precision) - single precision floating point calculations;
- FP64 (Double Precision) - double precision floating point calculations.
It is the last type of calculation (FP64, or binary64 according to the IEEE 754 standard) that is most in demand in BOINC calculations and other scientific programs, since it allows you to operate with values in the range from ≈ 10 -308 to 10 308 with an accuracy of up to 15 digits (in decimal expression) after comma. However, the share of calculations with single precision is also very large and even sufficient for individual projects.
Modern video cards support general purpose computing technologies. Any video chip released after 2007 will do a better job of scientific calculations than a similarly priced central processing unit of a similar generation.
Architecture and speed of calculations like FP32 and FP64 of some video cardsIf earlier nVidia dominated this segment (including due to the release of Tesla accelerators of the same architecture), now more and more projects are looking towards AMD (ATI).
An example of executing BOINC projects on an AMD video cardThere are also projects that support acceleration on AMD video cards, but do not work with nVidia products.
BOINC projects with exclusive support for AMD graphics cardsPresumably, this is due to the fact that at a high speed of FP32 calculations, nVidia video cards based on the GK104 chip demonstrate a strong performance drop in FP64-type calculations. For example, if the Radeon HD 6930 performs calculations with double precision four times slower than with single precision (480 and 1920 gigaflops, respectively), then the GeForce GTX 680 is twenty-four (128 and 3090 gigaflops).
When "Last" Doesn't Mean "Best"
By itself, the use of more modern chips does not always mean an increase in speed. For example, the HD 6850 is not capable of double precision calculations, while the HD 5850 performs FP64 calculations at up to 418 gigaflops.
High-end video cards often have the performance of a multiprocessor server. For example, a video card with an HD 7970 chip contains 2048 cores combined into 32 blocks. Its theoretical performance is 3789 gigaflops in single precision and 947 gigaflops in double precision.
For comparison: the arithmetic part of the Core-i5 3570K processor provides 122 gigaflops (FP32) and 61 gigaflops (FP64) in turbo mode, and Intel HD Graphics 4000 gives an increase of another 147 gigaflops (FP32). The AMD A10-5800K has a calculated performance of the arithmetic part also at the level of 122 gigaflops (FP32) and 61 gigaflops (FP64), but the Radeon HD 7660D video core provides a fourfold increase - by 614 gigaflops (FP32).
Now it definitely makes sense to choose those projects that can be calculated using video cards or at least integrated graphics cores. However, between the theoretical performance limit and the practically achievable result, there is often a difference not by percentages, but by several times. It is due to the qualifications of programmers and optimization limits for each specific task.
For example, project assignments [email protected] on a computer with a Core-i3 2100 and a Radeon HD 6850 video card, they are calculated at a speed of 49.5 gigaflops, and [email protected]- 71.4 gigaflops with the same settings. Both projects declare support for acceleration of calculations by AMD video cards.
Electricity costs
An important point is the energy efficiency of different systems. Now it is from 1 to 20 gigaflops per watt, and this parameter directly affects how your electricity bill increases. From the point of view of efficiency, not only specialized accelerators and server solutions are attractive, but also APUs and individual mass-produced video chips. AMD's most interesting chips are located closer to the beginning of the price range, while nVidia's are mostly in its upper third.
For example, the low-cost Radeon HD 7770 has an efficiency of 16 gigaflops per watt in FP32 calculations, but its use for FP64 calculations is unjustified - here it will show a result of only 1.0 gigaflops per watt. Even the HD 4750 chip, released in 2009, had a much better figure - 2.5 gigaflops per watt.
The GeForce GTX 660Ti belongs to the upper end of the mid-range price range, but uses energy more efficiently: 16.4 gigaflops per watt in FP32 calculations and 2.1 gigaflops per watt in double precision calculations.
If (theoretically) keep BOINC in constant processing mode and do not turn off the computer, then with a power consumption of ≈ 200 W, it will add less than 150 kWh per month to the meter readings. In a typical usage scenario (when BOINC is active only when idle, and the computer is running for several hours a day and turned off at night), less than ten kilowatt-hours.
Perhaps these small expenses in support of the research teams of the world's leading universities will seem to many more justified than transferring donations to dubious charities.
This is a short version of the BOINC client settings for those who do not like to read a lot and immediately "rush into battle". Despite the abundance of options and capabilities of BOINC, in most cases, you can leave all the default values, really important options that everyone needs to configure, in my opinion, there are only 3, which we will now configure in just 4 steps. Screenshots are from the English version of the client 6.12.26, the names of the options are given in English and duplicated in parentheses by the Russian translation in its current version (the translation may differ slightly from version to version).
1. By default, BOINC Manager works in "simplified" mode, in which many necessary functions are not available. Therefore, if you have not done this before, then switch to the "advanced" mode "(Advanced View) by pressing the appropriate button:
2. Open the BOINC Manager settings:
In versions of the program starting from 6.12.xx, they are located in the menu item Tools --> Computing preferences (Service --> Client Settings)
In older versions, in particular 6.10.xx, in the Advanced --> Preferences item (Advanced --> Settings)
3. On the "processor usage" tab, set the following options:
"While processor usage is less than XX percent" = 0
This option completely stops calculations when the processor is loaded by user programs (which means any programs other than BOINC itself, including those running in the background) more than XX%. 0% means ignore CPU usage by other programs. It is recommended to set exactly 0, because. calculations are carried out with the lowest priority and do not interfere with the work of other programs under normal conditions (because computing resources are dynamically redistributed in their favor by means of the OS). Or you can set your own value, for example 80-90% (i.e. pause the calculation only when the load is heavy). The default value (25%) is no good - calculations will stop at the slightest user activity, as well as, for example, starting a background antivirus scan or parallel use [email protected].
In case you want to use only a part of the processors/cores available in the computer for BOINC calculations, set it in the following option: "On miltiprocessor systems, use at most XX % of the processors" (In multiprocessor systems, use XX % of the total number of processors) . The value is given in %, i.e. to use only 3 cores on a 4-core processor, set it to 75%, 1 core on a 2-core processor - 50%, etc.
On multi-core/multi-threaded processors, this method is also preferable for throttling/reducing the load (e.g. in case of overheating) compared to "use at most XX % CPU time" (Use no more than XX % of each processor time), because it provides a more "smooth" load on the processor and additionally saves RAM (by reducing the number of processes/computing modules running in parallel).
If it is supposed to use all cores, then we simply leave the default value, i.e. 100%.
4. On the tab "network usage" (network settings) in the column "Additional work buffer XX days" (Additional buffer XX days) set the size of the job cache:
A large supply of tasks will allow you to avoid downtime in case of prolonged problems with the Internet or failures on the project servers. Recommended values are from 2 to 8 days. In this case, the more regularly the computer works, the more you can set the size of the job cache. Regularity means the mode of operation of the computer, ideal regularity is work in 24/7 mode (24 hours a day, 7 days a week) or every day, for example, exactly 10 hours. Office computers (about 9/5 mode) can also be considered fairly regular. But if the computer turns on sporadically - not every day and the duration of work is different every day, then it is undesirable to set a large job cache, because. the client will make big mistakes in calculating the required number of tasks (in pieces) to ensure the stock you specified (in days) and as a result, more tasks may be downloaded than you are able to complete on time. I do not recommend using the "Load tasks for XX days" item to set the cache size, because in addition to the cache size, it also sets the frequency of connecting to the project servers (in the original, this item is called "Connect about every XX days") and, by setting a large value there, we get a situation that the client will wait for several days before sending the counted tasks to the server and get a portion of fresh ones.
If you are interested in a more detailed setting or a description of some options, then we read the full BOINC setting option located here: (there will be a link when this full version is ready)
