Hi Pierre-Arnaud,
Yes, I created a new partition without any index attributes. I too am
lazy.. :)
Here's my 2 cents worth about - should Client or Server side do the
ordering.
1. If the Studio or Client side would to do the ordering, that would
mean all clients must implement the same piece of logic over and over
again. That may cause future support issues.
2. If the Server side do the ordering then what happen if Studio is used
against other Ldap Servers that does not order an exported entries?
For me now, it does not matter whether Client or Server do the
ordering, because the only tool i now use for Ldap development is
Apache DS and Studio and sometimes OpenLdap to confirm my development
works on it too.
Actually, we fixed the problem by reordering the ldif entries ourselves
manually. I thought it would be cool if Studio Ldif Editor has a feature
to re-order the ldif hierarchy.
Pierre-Arnaud Marcelot wrote:
Ok, I did more tests this morning.
With Emmanuel we figured out this was a problem of indexed attributes on our
testing partition.
My initial partition, where I'm getting the error does not have any index (I
was tool lazy to add some at the creation of the partition). This lack of
index, particularly on the 'objectClass' attribute makes Apache DS return
the search results in an unordered way.
For the purpose of the test I created the exact same partition but added
basic indices on it, I imported the very same data and the LDIF export was
correctly ordered.
Cheong Chung Onn, do you have the same thing? Did you create a new partition
with an empty list of indexed attributes?
Now, the question is: On the Studio side, Do we take for granted that search
results given from the LDAP Server are always correctly hierarchically
ordered (even if no RFC clearly specify it, just because it's the common way
of returning the results), or do we add additional code to verify and,
eventually, reorder the entries when writing the LDIF export file ?
Another question: Do we need to raise a JIRA for the problem we've face with
Apache DS when a partition does not have an index on the 'objectClass'
attribute ?
Regards,
Pierre-Arnaud
PS: A copy of this mail is attached to the JIRA
(DIRSTUDIO-448<https://issues.apache.org/jira/browse/DIRSTUDIO-448>),
to keep track of the discussion going on here.
On Thu, Jan 22, 2009 at 11:06 AM, Cheong Chung Onn <[email protected]
wrote:
Hi,
I have just created a JIRA
https://issues.apache.org/jira/browse/DIRSTUDIO-448
The export i did was always with scope "subtree". I just did another 3
round of tests using a new embedded DS Server. I made modifications to the
entries and exported ldif, each time the exported ldif file is in the
correct sequence now.
Stefan Seelmann wrote:
Hi,
I think the problem is when doing an export with scope "subtree". In that
case Studio just sends one search request with subtree scope and writes the
entries in the same order as received to the LDIF.
I think reordering them in Studio doesn't make sense, we can't hold them
all in memory.
An option would be to do multiple one-level scope searches.
Kind Regards,
Stefan
Quoting Emmanuel Lecharny <[email protected]>:
Cheong Chung Onn wrote:
We are using ADS 1.5.4, Studio 1.3.0. Java.
We created an embedded server in Studio then import the ldif file. Did
some modification and export it again. When we import this newly exported
ldif, we receive a bunch of error messages about missing object.
Please let me know if you require more info.
Ok, this was what we suspected... A clear bug somewhere.
Can you fill a JIRA with the informations you gave by mail ?
Eventually, providing the data you injected would help to reproduce the
problem, and determinate if it's a Studio or ADS problem. (of course,
if there are nothing confidential in it).
Many thanks !
--
--
cordialement, regards,
Emmanuel Lécharny
www.iktek.com
directory.apache.org
