Thank you very much!
Edit (I add more information and I answer to @user02814):
Thank you very much for your answer. I've actually read all that documentation, but the problem is that most were written years ago.
The conclusion I get is that the guard node is a very "controversial" issue and is being studied very well. I know it's a very serious issue, since that node is very special. Among other things, because it knows the IP of the Tor client. In addition, one of the methods to attack Tor is to know the first and last node and try to correlate the traffic.
The issue is that I have read that currently the Tor client chooses three entry guards, and after a certain time changes between them. But the question is: currently, how many entry guard does Tor choose and when does Tor change them?
I know it is a subject that is currently being studied and is prone to change in a short / medium period of time. In any case, in proposal 271 (created: 2016-07-11), called "Another algorithm for guard selection" (https://gitweb.torproject.org/torspec.git/tree/proposals/271-another-guard -selection.txt) it is said that "Tor clients choose a small number of guard nodes (currently 3)". However, proposal 236 (created: 2014-03-22), called "The move to a single guard node" proposes "switch to one guard per client. When this proposal becomes effective, clients will switch to using a single guard node".
So, my question is if someone knows how many "entry guards" the client chooses and how long it keeps them (maybe time varies depending on the circumstances, I do not know). Additionally, I would like to know if the "NumEntryGuards NUM" option affects the number of guard nodes that the client chooses. I have read the Tor manual but I do not quite understand what this option is for.
Edit (I answer to canonizing ironize and I give some more doubts):
So, the information that says Tor picks 3 'entry guards' is outdated today, right? Therefore, Tor, right now, chooses a single entry guard yes or yes, for these two reasons: first, because in the consesus is set "NumEntryGuards = 1" (although this could change). Second, because even if "NumEntryGuards = 1" did not appear in the consensus, the Tor source code specifies that the client of Tor have to choose only an entry guard, using "#define DEFAULT_N_GUARDS 1". Is what I'm saying correct?
On the other hand, in the Tor manual this is said about the "NumEntryGuards NUM" option: "If UseEntryGuards is set to 1, we will try to pick a total of NUM routers as long-term entries for our circuits". I do not quite understand this, could you explain it to me? I mean, I know that the "UseEntryGuards" option allows you to "turn off" the use of 'entry guards', but what relationship does it have with 'NumEntryGuards NUM'?
Additionally, I think there is an error in the Tor manual, precisely in the "NumEntryGuards NUM" option. It is said that "If NUM is 0, we try to learn the number from the NumEntryGuards consensus parameter, and default to 3 if the consensus parameter is not set (Default: 0)".
I think this should appears: "If NUM is 0, we try to learn the number from the NumEntryGuards consensus parameter, and default to 1 (ONE) if the consensus parameter is not set". If you confirm that this information is incorrect, I will open a 'ticket' to report it.