On Oct 4, 2007, at 10:38 AM, Matt Peterson wrote:

Alex,
 
I'm not sure that including two sets of operational parameters for calls to the partition is the right answer.  Doing this would allow partitions to be written that could not benefit from optimizations made by interceptors.  I like the interceptor paradigm because it encourages modularity and code reuse. I think that it would be wise to "encourage" partition authors to use interceptors appropriately and not embed into the partition what should be in an interceptor.
 
However, it would also be wise to allow users the option of not using certain interceptors with a particular partition. In other words, we need per-partition interceptor configuration.  I'm fairly focused on my current project, but I think it would be generally useful to make the interceptor configurations a property of the partition configuration (in server.xml).  This way users can select the set of interceptors that they'd like to use with a given partition.  Situations where interceptor functionality is required by a partition (such as normalization) could easily be detected in individual partition initialization code and result in the appropriate (graceful) warning or failure.
 

I think its clear that there are problems with the current "skip list" approach when people add custom interceptors that perhaps should only be used in some circumstances.  I've wondered if this problem could be solved by explicitly configuring the interceptor lists for each purpose in the spring xml rather than any of hardcoding the skips (current approach), hardcoding the "applies" (opposite and perhaps even worse approach) or building a logic engine to decide (I think this has in fact been tried by some projects).  I think jetspeed uses an approach similar to "configure lots of lists explicitly".

If this idea of configuring lots of interceptor lists in server.xml worked out it might be pretty easy for these to be specified per partition as well.

Not sure, but I think what I've described above would work for embedded implementations too...
 
--
Matt
 
P.S. - I'm still looking for an elegant way to implement interceptor bypass in the current 1.x codebase.  Still haven't found one that wouldn't also change the Interceptor interface (resulting in a cascade of changes).  Looking at a few more ideas before I give up.

Personally I think it's slightly easier to understand what is going on in the bigbang branch and if you pursue something like the "explicit lists" approach we could actually commit it without rewriting if you work against that moving target.

thanks
david jencks

 
On 10/4/07, Alex Karasulu <akarasulu@apache.org> wrote:
Also at some point we may include two sets of operation parameters.  One set is the
original set of parameters for the operation without any alterations taking place such as
those with normalization or other optimizations.  Another set will be the same parameters
tweaked by the system which should be applied by partitions. 

These are considerations for our bigbang branch which is refactoring the 1.5 line for a
2.0 release.  I think by providing the original request objects in the operation context as
well as the modified versions of them your problem will probably go away.

Still though feel welcome to follow this activity to make sure your needs are met as
we change the server's architecture.  Any recommendation or involvement would be
highly appreciated.

Alex


On 10/4/07, Matt Peterson <ickydude@gmail.com > wrote:
Ok.  I will.  Thanks.


On 10/3/07, Alex Karasulu <akarasulu@apache.org > wrote:
Yes I understand. If you would like to implement this bypass capability based on the target DN then feel
free to submit a patch.

Thanks,
Alex


On 10/3/07, Matt Peterson <ickydude@gmail.com > wrote:
Greetings,
 
I'm trying to write a custom partition that will be used as context partition, but I'd like to continue to use the Btree/Jdbm partition implementation for the system configuration partition.
 
With a little trial and error (mostly errors), I've found out that several methods in the JdbmPartition (used by the system configuration partition partition) will not work unless preceeded by a corresponding "fix-ups" provided by the interceptors.  For example, calls to to the methods in org.apache.directory.server.core.normalization.NormalizationService must be made, or BindDNs can not be resolved.  Unfortunately,  MyCustomPartition, doesn't like the way that LdapDNs, attributes and filters are normalized by org.apache.directory.server.core.normalization.NormalizationService. 
 
In other words, there appears to be more than at least one situation where a correlation between interceptors and partitions is necessary, but what we have (at least in Apache DS 1.0.x) is a design where all interceptors in the chain are called regardles of the partition that will eventually be invoked. To configure per-partition interception.
 
Perhaps what I'm looking for is a way that I can insert my interceptor first in the chain and then optionally skip other interceptors if I encounter a DN or base that I know will invoke my partition?  According to documentation for ApacheDS 1.0.x (http://directory.apache.org/apacheds/1.0/interceptors.html) the Nexus is called by a "special interceptor" at the end of the chain.  Therefore, it seems very difficult to skip interceptors in the chain without also elimintating the call to PartitionNexus ...
 
I suppose my real question is why ApacheDS wasn't designed so that interceptors could (optionally) be configured for association only with a certain suffix or partition.  This would allow the configuration flexibility that I would need -- to use both the JdbmPartition and my own customPartition without having to undo all of the normalization, schema, referral (,etc) stuff in my own customPartition that I didn't want done in the first place...
 
Does this make any sense?
 
--
Matt