Hi Brad, I’ve been watching this thread for a while, and you’ve finally managed to draw me in :)
On 12 Jul 2016, at 17:42, Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>> wrote: Guillaume, I'm still using Blueprint and have found Camel/SCR to be a pain to work with. It provides no tangible benefit if one is using Blueprint for routes and dependency injection anyway as it simply introduces one more way of configuring things. It was interesting to read the other day that Christian seems to have the same impression of the complexity of SCR. I remember when I first tried I thought it looked pretty cool but started running into problems. So what you’re comparing here isn’t really apples with apples. A long way back Camel provided a bunch of Spring XML sugar to make configuring Camel simpler. This was obviously a good thing because setting up Camel was (and is) hard. To this day people still use the Spring syntax for building their Camel routes. OSGi blueprint was effectively a move by Spring to standardise the Spring DM container, and so unsurprisingly blueprint looks a lot like Spring. Some people did the work to port the Camel Spring configuration to OSGi blueprint, and that’s why Camel is easy to use in blueprint. If someone actually spent some time putting together a nice integration with SCR then I’m sure that using SCR would be at least as easy as blueprint. The problem here is that relatively few SCR developers/users use Camel, and the ones that do are just told to “use blueprint”. That DS is in its second generation and only now getting around to transactions is telling. Either it has reached its natural boundaries and is now over-reaching or wasn’t full thought out. DS is actually working on its fifth release, and transactions are nowhere to be seen. You may be referring to the Transaction Control specification, which is separate from DS. They can be used together very effectively, but you could equally use Transaction Control with blueprint. DS is actually one of the “good citizens” of the DI world in that it deliberately does less in order to do it well. There’s no dependency proxying, no aspects, just the code that you wrote injected with some other code that someone wrote. To me it's a component and service bootstrapping mechanism which represents a small portion of the world I work in - transforms, routing, EIPs, etc. I've no reason to embrace it or deny it unless it either makes my job much easier or I can't live without it. So far adding Camel SCR and DS into the middleware just results in one more thing I have to deal with. This is a perfectly acceptable viewpoint. If the fundamental limitations of the blueprint model aren’t a problem for you then switching right now is almost certainly unnecessary. I think Blueprint works well these days and has come a long way in the past 3 years. The Aries team is to be commended for some great work. Aries Blueprint has had a lot of extensions and improvements over the last three years. Sadly the same cannot be said for the specifications or other implementations. Aries Blueprint is very much the last implementation standing, and there has been no effort to standardise the new features (or even to try fixing the problems with the original standard). The set of RFPs/RFCs for blueprint that have been sitting idle in the OSGi Alliance is very telling. As far as Aries blueprint is concerned, the main reason that it is still alive seems to be the fact that it was included in Karaf, and that Karaf provides Camel integration alongside it. Even Karaf itself is starting to move to use DS internally, offering blueprint as something for applications to use. I’ve been surprised by the near religious zeal of some of the DS advocates. Most OSGi developers I know (myself included) who really start to use DS consider it to be roughly equivalent to magic. The fact that the model can be as simple as it is and yet still flexible, correct and safe is both surprising and pleasing. Moving back to “not DS” is usually pretty painful and reminds people why they love DS so much. I'll be interested in seeing the DS semantics and proxies for CDI. Heh. Proxies are another technology that I don't care about one way or the other as long as things work well and don't require a lot of configuration. So it’s great if we can get rid of proxies but not so great if I now have to trade that off for configuration of start up order on services to make sure everything is running before Camel routes come up. Actually, this is one of the places where DS really shines. If you write a DS component properly (i.e. without trying to dig out of the DS lifecycle) then startup ordering ceases to be an issue. Again, someone with a little time and expertise would probably find that Camel + DS can be a really effective solution. The problem is finding the person who has the time, expertise and inclination… Tim Ward Apache Aries PMC Member, OSGi IoT Expert Group Chair, Author Enterprise OSGi in Action timothyjw...@apache.org<mailto:timothyjw...@apache.org> On Tue, Jul 12, 2016 at 10:57 AM, Guillaume Nodet <gno...@apache.org<mailto:gno...@apache.org>> wrote: I can happily review a patch if you're fancy writing one... And I disagree with the 'blueprint is dead and nobody cares about it anymore'. What's dead is the blueprint osgi specification for blueprint, not the Aries Blueprint project. I've recently added a bunch of important features related to spring in blueprint. DS also has some drawbacks as it's not extensible at all : this is leading the OSGi Alliance to write a new spec for transaction support !!! I think the CDI+DS extension I've been working on those past weeks could bring the best of both world : strong DS semantics for the OSGi bits, but extensibility and support for proxies provided by CDI ;-) 2016-07-12 17:24 GMT+02:00 Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>>: David, I'm all for multilocation support in blueprint. Can't wait for it. But it sounds like your saying blueprint is dead and nobody cares about it anymore so it doesn't seem likely to happen anytime soon. It certainly wouldn't be relevant to Fuse which uses R4 in any case. On Tue, Jul 12, 2016 at 10:21 AM, Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>> wrote: The features file can have statements like this: <configfile finalname="/etc/com.foo.cfg" override="true">mvn:com.confignuration/myrpoject/${project.version}/cfg/com.foo</configfile> <configfile finalname="/etc/com.bar.cfg" override="true">mvn:comconfiguration/myproect/${project.version}/cfg/com.bar</configfile> ....etc.... That's off the top of my head so take it with a grain of salt for syntax. When you run the features install it will overwrite the files in the etc directory with the ones in the maven bundle which have now been updated. So instead of modifying configuration files in the etc directory you modifying them in your Maven configuration project and recompile the bundle and then pull it from the repo in order to update the values. But you can still modify them in the etc if you wanted. You just have to make sure you have the cm properties set to reload. <cm:property-placeholder persistent-id="com.foo" update-strategy="reload"> On Tue, Jul 12, 2016 at 9:45 AM, Pablo Gómez Pérez <pablo.go...@faw.jku.at<mailto:pablo.go...@faw.jku.at>> wrote: Brad, if I understand your approach too that would lead to not being able to dynamically change common properties in a single config place during runtime, as the fill made by maven would be only done once at build time right? But at runtime I would need to that as David mention, still n times right? as a use case for instance, with blueprint:cm update-strategy configured as 'reload' in combination with felix-fileinstall as directory watcher, bundles are reloaded automatically so that when I modify at anytime during runtime a property e.g with just a text editor the bundle is initiated again with the new property values which is a quite nice feature best Pablo On 12/07/2016 12:31 AM, David Jencks wrote: I’d like to make sure I understand what you are doing here…. IIUC during the build of your project you are generating multiple configuration files with the same or similar content, and each of these is loaded into a configuration which is bound to a particular bundle location? So, at build time you can change all the duplicate properties at once but if you need to change them later you have to alter n (== number of duplicate configs) independently? Whereas if you had multi-location support (and possibly multi-pid support such as DS provides) you could share a single Configuration and change the property while the framework was running in one place? thanks david jencks On Jul 11, 2016, at 1:42 PM, Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>> wrote: Pablo, One possible solution to this problem that I'm currently looking at is using a configuration bundle along with my features bundle. In the configuration bundle I have all the cfg files and use properties placeholders ${value} to set the value for key/value. In the POM I load properties files using the Maven properties plugin and that lets me set a global set of properties values that can be used in filling in the cfg values. So if a port or URI is shared across a large number of them that automatically gets filled in. The features file can then specify the cfg files to install and what name to install them with. This gets rid of a lot of tedium and by using profiles I should be able to switch dev, test and production, and have the properties automatically set correctly. I'd like to modify this a bit so that dev, test and product cfg files are all created simultaneously and simply installed in different directories inside the configuration bundle. Then by using different features installs I can easily switch between the different configurations without having to tediously edit each configuration file. Brad On Sat, Jul 9, 2016 at 11:32 AM, Matt Sicker <boa...@gmail.com<mailto:boa...@gmail.com>> wrote: Does Camel/Fuse even support DS? I haven't found any documentation saying otherwise. I've only found camel-scr which uses Felix-specific annotations and not DS. On 7 July 2016 at 14:32, Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>> wrote: David, That is some pretty extreme and wild speculation alright. How does one use blueprint to not use OSGi appropriately? In the 5 years I've been consulting with Fuse/Karaf/OSGi and going to various clients not one of them used or uses DS. Not one. They all use bundles, services, and Camel with blueprint. The last time I worked with DS I didn't find it provided any serious advantage and added another layer that I'd have to teach my clients. Not that I wouldn't consider it or use it if I found a real advantage but I haven't. Red Hat is still shipping Karaf 2.x with Fuse so it is still in OSGi 4.x land much less 5 or 6. So for Camel are you using the Java DSL? Brad On Thu, Jul 7, 2016 at 1:56 PM, David Jencks <david_jen...@yahoo.com<mailto:david_jen...@yahoo.com>> wrote: I don’t think karaf is at osgi R4.2 any more, I suggest you look at the osgi R5 or R6 config admin spec for “multi location”. You guys might be using blueprint every day, but there is no OSGI spec work to keep it up to date or even specify obviously necessary features such as config admin integration. If blueprint is so great why aren’t the proponents keeping the spec related to current OSGI? This is a part of my, admittedly extreme, theory that use of blueprint is related to not wanting to make the app actually use osgi appropriately. And, the project I work on every day uses DS exclusively and still finds plenty of ways to abuse osgi in all sorts of inventive ways :-) david jencks On Jul 7, 2016, at 11:11 AM, Johan Edstrom <seij...@gmail.com<mailto:seij...@gmail.com>> wrote: It is in here; https://osgi.org/javadoc/r4v42/org/osgi/service/cm/ConfigurationAdmin.html A bundle is in aries bound to the pid. So it is actually working as expected, bit of a hassle since spring-dm allowed it. And yes selling DS into “regular" organizations is about as easy as selling snow in Alaska. /je On Jul 7, 2016, at 12:00 PM, Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>> wrote: David, You live in a very different world than I do. In all the consulting I do with Fuse/karaf blueprint is used almost exclusively. I understand DS and its uses but also its limits and overhead. It's like telling me one should only use Camel Java DSL. That may be one's perspective but that isn't everyone's. Brad On Thu, Jul 7, 2016 at 12:53 PM, David Jencks <david_jen...@yahoo.com<mailto:david_jen...@yahoo.com>> wrote: IMNSHO blueprint is only really plausible if you have a large amount of Spring based code and you need to convert it to be sort of osgi-compatible really quickly without understanding osgi or the code. Otherwise taking the time to understand DS and use it is much more satisfactory. DS provides this configuration override ability with support for multiple pids, although only one of the pids can turn out to be a factory configuration. There’s no obvious way of correlating factory configurations, so this restriction makes some sense. I don’t think there really are any blueprint folks. The cm stuff, while obviously required to make the spec remotely plausible, hasn’t made it into the spec in the many many years it’s been sitting around. david jencks On Jul 7, 2016, at 10:41 AM, Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>> wrote: If I were to sit down with the blueprint folks today to create a wish list one thing I'd like to see is for an ability to have a configuration hierarchy specified with parent/child relationships much like one has in Maven. Have a base configuration file and be able to have another cfg file specify that one as its parent. Override properties or add them to the child. When the configuration admin fires up it would read up the chain and construct the properties. On Thu, Jul 7, 2016 at 12:37 PM, Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>> wrote: Ray, If I understand your question right the answer is the Aries extension is referencing configuration. In karaf/fuse for example the following: <cm:property-placeholder persistent-id="com.my.foo" update-strategy="reload"> will load properties from etc/com.my.foo.cfg Installing that file is done either manually or by use of a features file. Whenever I've attempted to use the PID in more than one bundle it has failed and I don't think it is permitted. That's a problem I think and something that should be fixed through some other configuration management mechanism. Making microservices that might share common properties, for example, becomes problematic in that regard and I've resorted to using my own OSGi services to overcome that problem. Brad On Thu, Jul 7, 2016 at 10:46 AM, Raymond Auge <raymond.a...@liferay.com<mailto:raymond.a...@liferay.com>> wrote: Ok, so after a brief review the cm schema is an Aries extension and it doesn't appear to support the location binding. However, it's unclear to me whether this extension is creating the configuration or merely referencing one from outside. Any Aries gurus can answer that? - Ray On Thu, Jul 7, 2016 at 11:29 AM, David Jencks <david_jen...@yahoo.com<mailto:david_jen...@yahoo.com>> wrote: I’m not really familiar with blueprint cm but I’d expect that to indicate which pid to use to fetch the config from config admin and in the ... how to map configuration propertiething blueprint substitution knows about. Is that really instructions to create a new configuration and populate it with data (what a management agent does)? david jencks On Jul 7, 2016, at 8:19 AM, Raymond Auge <raymond.a...@liferay.com<mailto:raymond.a...@liferay.com>> wrote: David, I agree with everything you've said, however this looks like blueprint being the agent here: <cm:property-placeholder persistent-id="my.id<http://my.id/>" update-strategy="reload"> ... </cm:property-placeholder> - Ray On Thu, Jul 7, 2016 at 11:18 AM, David Jencks <david_jen...@yahoo.com<mailto:david_jen...@yahoo.com>> wrote: No, blueprint cm shouldn’t really know about the multi-location. The management agent that is creating the configuration should be setting the bundle location to the multi-location ”?”. david jencks On Jul 7, 2016, at 8:12 AM, Pablo Gómez Pérez <pablo.go...@faw.jku.at<mailto:pablo.go...@faw.jku.at>> wrote: I see and would it possible to configure which method is invoked from Blueprint? This is how I do it: <cm:property-placeholder persistent-id="my.id<http://my.id/>" update-strategy="reload"> ... </cm:property-placeholder> is there perhaps some blueprint property where I can tune the second argument in the createFactoryConfiguration? Because it looks like the fact of using config admin through blueprint binds the PID to the first bundle using it best Pablo On 07/07/2016 4:41 PM, Raymond Auge wrote: As long as configurations are not bound to a bundle they can be used by any bundle. The exception clearly shows that the configuration is bound to a bundle. Creating an unbound configuration requires passing a "?" as the second arguments to getConfiguration/createFactoryConfiguration methods of CM. HTH, - Ray On Thu, Jul 7, 2016 at 10:24 AM, Brad Johnson <brad.john...@mediadriver.com<mailto:brad.john...@mediadriver.com>> wrote: I don't think that's possible. On Thu, Jul 7, 2016 at 8:51 AM, Pablo Gómez Pérez <pablo.go...@faw.jku.at<mailto:pablo.go...@faw.jku.at>> wrote: Hello All, Is it possible to use same config file from multiple bundles while using Config Admin with blueprint Blueprint? Because, I can't manage to do that, I get the following error: MESSAGE Cannot use configuration test.mybundle for [org.osgi.service.cm<http://org.osgi.service.cm/>.ManagedService, id=214, bundle=86/initial@reference:file:../plugin-1/<mailto:bundle=86/initial@reference:file:../plugin-1/>]: No visibility to configuration bound to initial@reference:file:../plugin-2/<mailto:initial@reference:file:../plugin-2/> I saw in this jira a bug opened: https://issues.jboss.org/browse/ENTESB-3959 However, I fear that this is a problem in the aries blueprint implementation as I'm not using KARAF nor FUSE, just a plain osgi container. Either that or I'm missing some blueprint configuration. I'm basically using blueprint:cm As a workaround I can make a config file per bundle that needs it.... As follows the versions and bundles that I'm using related to the container (Running on top of Equinox 3.11): ID|State |Level|Name 5|Active | 2|Apache Aries Whiteboard support for JMX DynamicMBean services (1.1.5)|1.1.5 6|Active | 2|Apache Aries JNDI Core (1.0.2)|1.0.2 13|Active | 3|Aries Remote Service Admin Topology Manager (1.9.0.SNAPSHOT)|1.9.0.SNAPSHOT 15|Active | 2|Aries JPA Container (1.0.2)|1.0.2 21|Active | 2|Apache Aries JNDI API (1.1.0)|1.1.0 25|Active | 3|Aries Remote Service Admin Discovery Gogo Commands (1.9.0.SNAPSHOT)|1.9.0.SNAPSHOT 27|Active | 2|Apache Aries Blueprint CM (1.0.7)|1.0.7 29|Active | 2|Apache Aries JMX Blueprint Core (1.1.5)|1.1.5 37|Active | 2|Apache Aries JNDI URL Handler (1.1.0)|1.1.0 42|Active | 2|Apache Aries JMX Core (1.1.5)|1.1.5 46|Active | 2|Apache Aries Blueprint Core (1.5.0)|1.5.0 47|Resolved | 4|Apache Aries Blueprint Core Compatiblity Fragment Bundle (1.0.0)|1.0.0 55|Active | 2|Apache Aries Util (1.1.1)|1.1.1 56|Active | 2|Aries JPA Container Managed Contexts (1.0.4)|1.0.4 59|Active | 2|Apache Aries Proxy API (1.0.1)|1.0.1 67|Active | 3|Aries Remote Service Admin Service Provider Interface (1.9.0.SNAPSHOT)|1.9.0.SNAPSHOT 71|Active | 2|Apache Aries Transaction Blueprint (1.1.1)|1.1.1 73|Active | 2|Aries JPA Container API (1.0.2)|1.0.2 77|Active | 2|Apache Aries JNDI Support for Legacy Runtimes (1.0.0)|1.0.0 88|Active | 2|Apache Aries JMX Blueprint API (1.1.5)|1.1.5 89|Active | 2|Apache Aries Transaction Manager (1.3.0)|1.3.0 94|Active | 3|Aries Remote Service Admin Discovery Config (1.9.0.SNAPSHOT)|1.9.0.SNAPSHOT 97|Active | 3|Aries Remote Service Admin provider TCP (1.9.0.SNAPSHOT)|1.9.0.SNAPSHOT 110|Active | 2|Apache Aries Blueprint Annotation API (1.0.1)|1.0.1 120|Active | 2|Apache Aries Transaction Blueprint (2.1.0)|2.1.0 123|Active | 2|Apache Aries JMX API (1.1.5)|1.1.5 130|Active | 2|Apache Aries Blueprint Annotation Impl (1.0.1)|1.0.1 132|Active | 3|Aries Remote Service Admin Discovery Zookeeper (1.9.0.SNAPSHOT)|1.9.0.SNAPSHOT 134|Active | 3|Aries Remote Service Admin Discovery Local (1.9.0.SNAPSHOT)|1.9.0.SNAPSHOT 138|Active | 3|Aries Remote Service Admin Core (1.9.0.SNAPSHOT)|1.9.0.SNAPSHOT 139|Active | 2|Apache Aries JNDI RMI Handler (1.0.0)|1.0.0 143|Active | 2|Apache Aries Proxy Service (1.0.4)|1.0.4 146|Active | 2|Apache Aries SPI Fly Dynamic Weaving Bundle (1.0.8)|1.0.8 147|Active | 2|Aries JPA Container blueprint integration for Aries blueprint (1.0.4)|1.0.4 11|Active | 4|Apache Felix File Install (3.5.4)|3.5.4 19|Active | 4|Apache Felix Gogo Shell (0.12.0)|0.12.0 57|Active | 4|Apache Felix Gogo Command (0.16.0)|0.16.0 104|Active | 4|Apache Felix Coordinator Service (1.0.2)|1.0.2 109|Active | 4|Apache Felix Gogo Runtime (0.16.2)|0.16.2 114|Active | 4|Apache Felix Web Management Console (1.2.8)|1.2.8 148|Active | 4|Apache Felix Configuration Admin Service (1.8.8)|1.8.8 0|Active | 0|OSGi System Bundle (3.11.0.v20160603-1336)|3.11.0.v20160603-1336 -- WARNING: Computer viruses can be transmitted via email. The recipient should check this email and any attachments for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. E-mail transmission cannot be guaranteed to be secure or error-free as information could be intercepted, corrupted, lost, destroyed, arrive late or incomplete, or contain viruses. The sender therefore does not accept liability for any errors or omissions in the contents of this message, which arise as a result of e-mail transmission. Warning: Although the company has taken reasonable precautions to ensure no viruses are present in this email, the company cannot accept responsibility for any loss or damage arising from the use of this email or attachments. -- Raymond Augé<http://www.liferay.com/web/raymond.auge/profile> (@rotty3000) Senior Software Architect Liferay, Inc.<http://www.liferay.com/> (@Liferay) Board Member & EEG Co-Chair, OSGi Alliance<http://osgi.org/> (@OSGiAlliance) -- Raymond Augé<http://www.liferay.com/web/raymond.auge/profile> (@rotty3000) Senior Software Architect Liferay, Inc.<http://www.liferay.com/> (@Liferay) Board Member & EEG Co-Chair, OSGi Alliance<http://osgi.org/> (@OSGiAlliance) -- Raymond Augé<http://www.liferay.com/web/raymond.auge/profile> (@rotty3000) Senior Software Architect Liferay, Inc.<http://www.liferay.com/> (@Liferay) Board Member & EEG Co-Chair, OSGi Alliance<http://osgi.org/> (@OSGiAlliance) -- Matt Sicker <boa...@gmail.com<mailto:boa...@gmail.com>> -- ------------------------ Guillaume Nodet ------------------------ Red Hat, Open Source Integration Email: gno...@redhat.com<mailto:gno...@redhat.com> Web: http://fusesource.com<http://fusesource.com/> Blog: http://gnodet.blogspot.com/
