Re: [jabberd2] digest-md5 breaks gssapi

2009-01-04 Thread Andreas Hofmeister

Hi,

I did some further testing and hacked up a little test program that  
simulates the client behavior ( using a completely different SASL  
implementation). Apparently the problem is at the client side (i.e  
Adium/libpurple) which seems not to cleanup properly after the first  
of its two authentication attempts.


So it seems I need to disable digets-md5 in my configurations when  
using gssapi for the timi being.


Ciao
  Andi

--
To unsubscribe send a mail to jabberd2+unsubscr...@lists.xiaoka.com



[jabberd2] digest-md5 breaks gssapi

2009-01-03 Thread Andreas Hofmeister

Hi,

I tried several authentication scenarios with jabberd 2.2.4 (+gsasl  
0.2.29) and found that there seems to be a conflict between digest- 
md5 and gssapi. I tried


- gssapi + plain (with ldapfull and PAM), which works
- digest-md5 + plain with ldapfull, which works too.

Any combination of of gssapi + digest-md5 however causes  
authentication to fail with either mechanism. One of the clients  
(Adium) reported it had received a malformed challenge.


My suspicion is that both mechanisms trash each others idea of the  
authentication realm, but that is not very deeply founded.


Any idea what to try or look at next ?

Ciao
  Andi

--
To unsubscribe send a mail to jabberd2+unsubscr...@lists.xiaoka.com



Re: [jabberd2] digest-md5 breaks gssapi

2009-01-03 Thread Tomasz Sterna
Dnia 2009-01-03, sob o godzinie 10:47 +0100, Andreas Hofmeister pisze:
 My suspicion is that both mechanisms trash each others idea of the  
 authentication realm, but that is not very deeply founded.
 
 Any idea what to try or look at next ?

1. Compile jabberd with debugging, then run c2s -D and look for signs of
the problem.

2. Look at the decoded SASL exchange (you may find it in debug messages)
for incompatibilities.

-- 



-- 
To unsubscribe send a mail to jabberd2+unsubscr...@lists.xiaoka.com




Re: [jabberd2] digest-md5 breaks gssapi

2009-01-03 Thread Andreas Hofmeister


Am 03.01.2009 um 14:31 schrieb Tomasz Sterna:


Dnia 2009-01-03, sob o godzinie 10:47 +0100, Andreas Hofmeister pisze:

My suspicion is that both mechanisms trash each others idea of the
authentication realm, but that is not very deeply founded.

Any idea what to try or look at next ?


1. Compile jabberd with debugging, then run c2s -D and look for  
signs of

the problem.

2. Look at the decoded SASL exchange (you may find it in debug  
messages)

for incompatibilities.


Ok, I logged an attempt with GSSAPI only and one with GSSAPI + Digest- 
MD5 enabled.


In the GSSAPI only case, Adium uses GSSAPI only which succeeds.

In the second case, Adium first tries to authenticate using Digest- 
MD5 and as it has no password, this first attempt fails. It then  
tries GSSAPI immediately without reseting the connection, but  
terminates the connection the first challenge.


When I compare the successful GSSAPI attempt with the failing one,  
both look quite similar except that in the failing case the first  
challenge is 272 instead of 276 Bytes.


Ciao
  Andi

--
To unsubscribe send a mail to jabberd2+unsubscr...@lists.xiaoka.com