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

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

Alistair Young wrote:
> Jon, if your suspicions are correct and that the thread doesn't have  
> time to get ready, then adding more sleeps would presumably make it  
> worse? What happens if you remove them? Were they there to cure the  
> original problem which went away itself?
>   
I find that the longer those sleeps are the less often the exceptions 
occur - my interpretation is that this allows the API's background 
threads to catch up with the API calls but it is hard to trace this kind 
of problem because the exceptions occur in the background thread are 
caught in the background thread and their stack trace refers to that 
thread. With a zero delay the exceptions come thick and fast.

There's a limit to how much time I can put into studying this. In my 
personal opinion a good stable implementation of a network protocol 
should always report errors up to the application layer and not simply 
spew out stack traces on stderr and in the worst case silently freeze 
up. I have lost confidence in the Novell code. (By the way , can anyone 
find a bug reporting web site?)

Another good reason to switch to the Sun LDAP provider is that it 
conforms to JNDI which is the directory equivalent of JDBC. That means 
if we ever experience problems with the Sun provider we can drop in 
another one without wasting time reprogramming to a new API - you just 
drop a jar file into the lib directory and edit the config file.

Anyway, I realise that I'm in a minority of 1. The Leeds specific 
authenticator is now based on JNDI and I'm not going to switch it back 
to the Novell API because I need to get on with the next job. However, 
I'd be perfectly happy if someone else wants to rewrite it for release.

Jon