Re: [API] [ApacheDS] Problems with extended operations
On 2 août 2013, at 20:50, Kiran Ayyagari wrote: > On Fri, Aug 2, 2013 at 10:43 PM, Pierre-Arnaud Marcelot > wrote: > Hi, > > Recently, a user raised an issue on the LDAP API indicating that result codes > from extended operations were lost. > See https://issues.apache.org/jira/browse/DIRAPI-151. > > I took some time to debug the issue and it turns out it's true and it's even > bigger than that, the response is simply not parsed correctly and internally > in the API the object we return is just a newly created response for the > given extended operation. > The result code returned from the server is completely forgotten, same thing > for the specific values included in the extended operation response. > > On the server, Emmanuel also found that responses were not correctly encoded. > > We're currently fixing that. > > This issue lead us to have a look at how the extended operations are loaded > and I took a chance at simplifying a bit (which will help resolving the > issue) the list of factories we need to append in the command line. > We had properties like: > - default.controls > - extra.controls > - default.extendedOperation.requests > - default.extendedOperation.responses > - extra.extendedOperations > > In the end, those default/extra distinctions didn't mean much and we needed > to have them all the time. > Also, the titles of the properties like 'default.extendedOperation.responses' > were misleading as it was only intended for some specific kind of unsolicited > response (having no request). > > I simplified it to two simple properties: > - apacheds.controls > - apacheds.extendedOperations > > +1 for the move, but please support the old properties as well to keep this > change backward compatible > cause these properties(old and new as well) are not so explicit and many > users may not think of these > while debugging an issue, so backward compatibility saves them a lot of time > when they update their > server libraries (but using the old scripts/classes in their applications to > start the server) Sure, good idea. Regards, Pierre-Arnaud > All tests, scripts and installers have been updated to reflect this change, > so everything should be covered. > > Thoughts? > > Regards, > Pierre-Arnaud > > > > -- > Kiran Ayyagari > http://keydap.com
Re: [API] [ApacheDS] Problems with extended operations
On Fri, Aug 2, 2013 at 10:43 PM, Pierre-Arnaud Marcelot wrote: > Hi, > > Recently, a user raised an issue on the LDAP API indicating that result > codes from extended operations were lost. > See https://issues.apache.org/jira/browse/DIRAPI-151. > > I took some time to debug the issue and it turns out it's true and it's > even bigger than that, the response is simply not parsed correctly and > internally in the API the object we return is just a newly created response > for the given extended operation. > The result code returned from the server is completely forgotten, same > thing for the specific values included in the extended operation response. > > On the server, Emmanuel also found that responses were not correctly > encoded. > > We're currently fixing that. > > This issue lead us to have a look at how the extended operations are > loaded and I took a chance at simplifying a bit (which will help resolving > the issue) the list of factories we need to append in the command line. > We had properties like: > - default.controls > - extra.controls > - default.extendedOperation.requests > - default.extendedOperation.responses > - extra.extendedOperations > > In the end, those default/extra distinctions didn't mean much and we > needed to have them all the time. > Also, the titles of the properties like > 'default.extendedOperation.responses' were misleading as it was only > intended for some specific kind of unsolicited response (having no request). > > I simplified it to two simple properties: > - apacheds.controls > - apacheds.extendedOperations > > +1 for the move, but please support the old properties as well to keep this change backward compatible cause these properties(old and new as well) are not so explicit and many users may not think of these while debugging an issue, so backward compatibility saves them a lot of time when they update their server libraries (but using the old scripts/classes in their applications to start the server) > All tests, scripts and installers have been updated to reflect this > change, so everything should be covered. > > Thoughts? > > Regards, > Pierre-Arnaud > -- Kiran Ayyagari http://keydap.com
[API] [ApacheDS] Problems with extended operations
Hi, Recently, a user raised an issue on the LDAP API indicating that result codes from extended operations were lost. See https://issues.apache.org/jira/browse/DIRAPI-151. I took some time to debug the issue and it turns out it's true and it's even bigger than that, the response is simply not parsed correctly and internally in the API the object we return is just a newly created response for the given extended operation. The result code returned from the server is completely forgotten, same thing for the specific values included in the extended operation response. On the server, Emmanuel also found that responses were not correctly encoded. We're currently fixing that. This issue lead us to have a look at how the extended operations are loaded and I took a chance at simplifying a bit (which will help resolving the issue) the list of factories we need to append in the command line. We had properties like: - default.controls - extra.controls - default.extendedOperation.requests - default.extendedOperation.responses - extra.extendedOperations In the end, those default/extra distinctions didn't mean much and we needed to have them all the time. Also, the titles of the properties like 'default.extendedOperation.responses' were misleading as it was only intended for some specific kind of unsolicited response (having no request). I simplified it to two simple properties: - apacheds.controls - apacheds.extendedOperations All tests, scripts and installers have been updated to reflect this change, so everything should be covered. Thoughts? Regards, Pierre-Arnaud
