It is late but let my creative comments fly, no offense Chris is a very helpful guy, I am his fan(even if the rest would contradict).

Please consider that to me and perhaps others embedding ApacheDS, having an xml to edit is a great thing!

Loading things into a DIT is less pleasant:
- if it is not LDIF, and I have to start bootstrap the thing to create my own config. Like the situation of the schema system right now where I have to modify the build and work in your source tree to create my custom bootstrap package.

- if LDIF... it is still a weird syntax to use for IOC assembly. I would like to modify, add my custom interceptors, and components. One of the thing that is less fortunate about the Store procedures and triggers that I needed serialized class files, which is hassle for embedding. For a build I would have to setup, config, dump config, teardown.

- I would not like to put/encode xml into an LDIF file.....

I realize that admin tools, vs embedding is in a constant battle.
embedding needs configuration without actually running the DS.
admin tool want to configure a bare instance, any possible aspects after the DS is started.

Both are legit cases, with very different objectives.

I like DS because it is a Lego toolkit glued together by the IOC, Please do not take that way from me. DS is great because it is embeddable easy to customize keep that aspect alive!

Aron

Chris Custine wrote:
This is basically a response to some of the other threads regarding server.xml and Config in DIT, but I don't want to derail those threads if this turns out to be a stupid idea.

I have been thinking about this for a while, and I have to admit that I am one of the guys that likes Spring and the xml config files. Because of that I have been thinking about possible interim steps so that we can get a good grip on the needs and wants of the users while still solving some of our internal problems that we want to address in the short term. Based on the recent threads about this topic, I get the distinct feeling that we might be underestimating the affect this subject has as far as user impact and user preferences and stand a good chance of irritating some people.

My latest idea seems really obvious the more I think about it... For the time being, why don't we just move towards storing the server.xml in its entirety as a string attribute under ou=system somewhere and restructure the startup sequence to properly read and load the Spring context from there? This sounds crazy, but bear with me for a moment... This would give us the ability to "configure in DIT" so to speak, but would also expose some really interesting options for remote configuration, like modifying the current Apache DS Configuration plugin that Pierre-Arnaud has already written to just read from and save to the server it is currently connected to. We could also do an interceptor or something similar on the server to write the file out to disk after a remote edit and allow a startup option or quick command line script to load a new file after you edit it in vi or similar. This CLI could even put the data directly to the JDBM tables so that you can make edits without the server running.

I have a couple of reasons for bringing this to the table. First of all, I am one of those dirty, sadistic perverts that likes editing xml files by hand as opposed to many other forms of config... xml is like a second language to me :-) Second and most importantly for ApacheDS, is that an approach like this will give us a great short term benefit of remote config and admin capability, without all of the work. The server config editor that PAM wrote looks fantastic to me and (hopefully) we can just extend the concept and hack it to do what is needed for this without re-writing all of it. This way we can do the Config in DIT in an incremental fashion and possibly save some grief that we may encounter later. We will also be able to move on more quickly to more serious tasks and implement high visibility features.

I am sure there are some technicalities that may be obstacles to this idealistic approach, but I have had worse ideas, so I thought I would bring it up and present it. What do you guys think?

Thanks,
Chris

Reply via email to