Re: [Bodington-developers] More LDAP Fun and Games

[Jon M. <jo...@te...>]

O.K. the original problem we had was pinned down and resolved. (Would 
have told you earlier if I hadn't been distracted by just trying to get 
message through at all.) By that time I had rewritten the code to use 
the JNDI API but I'm pretty sure that we could have got the Novell code 
to stop throwing the exceptions we observed.  However, I'm still alarmed 
by the results of the test we ran outside of Bodington....

The loop was connecting, making a search for the DN of a record matching 
a given user name and disconnecting. I put Thread.currentThread.sleep() 
between every API call and at the end of the main loop. Although a delay 
of a whole second between API calls reduced the rate of exceptions to 
less than one in several hundred cycles they didn't go away completely. 
The thing that really bothers me is that none of the exceptions thrown 
resulted in a meaningful error condition or result message in the 
application itself - just nasty looking stack traces uncontrollably 
pumped out to standard error by the Novell API worker threads. That sort 
of thing should NEVER happen in a well written API - if there is an 
error condition it ought to be signaled properly to the application 
level. The test was carried out using my laptop as the client (plugged 
into the campus ethernet) and about three of the Active Directory 
servers on campus. I can supply the test program to anyone who would 
like to give a go.

To repeat the key finding since my last posting - the main problem that 
prompted the posting has been resolved and the Novell API is not to 
blame but we have an inexplicable problem occurring in a test scenario 
which reduces my confidence in the Novell solution.  So far the JNDI 
distributed with JDK 1.5 works fine with Active Directory although I 
haven't yet subjected it to the testing under load that we tried on the 
Novell API.