summaryrefslogtreecommitdiffstats
path: root/tdeio/kpasswdserver/DESIGN
blob: 6a128f9dfb73fd15722d35a6c32bd29d26758749 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
Sequence numbers
================
The idea is that whenever the user is queried for a password this 
login/pw combination gets a seq-nr. When  a slave needs a  login/pw 
it asks kpasswdserver and sends along the last seqnr it received. If 
this seqnr is older (lower) than the seq nr of the login/pw 
combination stored in kpasswdserver then appearantly the user has 
already been prompted for a new login/pw combination since the last 
time this slave asked for a login/pw and therefor it is not necessary 
to prompt the user again but kpassword will send the io-slave this 
new login/pw combination. If this new combination fails as well the 
user is being prompted for a new login/pw combo since the one stored 
in kpasswdserver doesn't work.

Let me try to draw the situation I had in mind when writing this:

Slave1		Slave2			kpasswdserver
Asks for auth
					asks user for login/pw (1)
sends login/pw (1) to ftp site
		Asks for auth
					sends back login/pw (1)
		sends login/pw (1) to ftp site
gets login error,
asks for new auth
sends along seq.nr 1
					seq.nr 1 == (1) -->
					asks user for new login/pw (2)
sends login/pw (2) to ftp site
		gets login error,
		asks for new auth
		sends along seq.nr 1
					seq.nr 1 < (2) -->
					don't ask user for new login/pw
					but send back login/pw (2) without asking
		sends login/pw (2) to ftp site


Actually, I had mostly http in mind, and not so much ftp. In http you 
typically try without password first, and only when you get an 
authentication error you ask for a password. The above scenario is 
then suddenly a lot more common than with ftp because it can happen 
that you have 4 requests /io-slaves who alll discover at about the 
same time that they need to have authentication credentials. The 
above scenario (and the seq. nrs) is to prevent that you get 4 login 
dialogs in such case.

Now the assumption in this all, looking back on it, seems to be that 
when you ask for the same auth credentials twice in a row, it must be 
that the credentials issued the first time where wrong, and you will 
be prompted again. But if the user goes to ftp-site1, then 
ftp-site2 and then back to ftp-site1 again, the credentials for ftp-site1
are still valid. This is why we reset the seq.nr stored in the io-slave
to 0 whenever the io-slave switches hosts (or logins).

Waldo Bastian <[email protected]>