Re: PAX:WEB How many war bundles do pax-web support?
Wasn't history preserved when we moved to Github? I don't see Alin's work pre-2010... Niclas On Mon, Oct 3, 2016 at 12:35 AM, 'Achim Nierbeck' via OPS4J < ops4j@googlegroups.com> wrote: > About a person year, with the prerequisite that the current implementation > is well known. > > 1) Not really, certain people working on the code are paid to do so. > 2) Nope, it's an open source project most people working on this do this > in their private time. > Sometimes those people (like me) don't even work on any related stuf > anymore. > 3) Hard to tell ... as we have people which get engaged for certain topics > which move off after a certain amount of time > Here's a complete list of recent developers [1] > > > regards, Achim > > [1] - https://github.com/ops4j/org.ops4j.pax.web/graphs/contributors > > 2016-10-02 18:20 GMT+02:00 Pavel Kastornyy: > >> Achim, thank you for the information. So about one person year >> if I understand you right. >> >> Could you also shortly answer the following questions: >> >> 1) is there any financial help from any companies? >> 2) has the community tried to draw investments into the product? >> 3) how many active developers are there at this time? >> >> Best regards, >> >> >> >> On 02.10.2016 18:47, 'Achim Nierbeck' via OPS4J wrote: >> >>> wow, that's a tough one ... :D >>> >>> if you take a look at Openhub [1] ... it'll tell you it took about 57 >>> years >>> [2], or at least it's the amount of work worth it ;) >>> >>> Anyway it's hard to estimate as I've spent the last 6 years improving on >>> it. Never the less if I would work on it full time 8h day >>> with the knowledge I have right now. One person could redo it within >>> maybe >>> a year. But it's a wild guess ... might be faster, might take longer ... >>> Of course I would expect it to have the same features and possibilities >>> it >>> has which makes it different to other implementations of the HttpService >>> etc. Right now we have about 500 unit and integration tests running with >>> every build [3]. >>> That functionality I would expect to be available after the re-write ;) >>> >>> BUT, the nice thing is. We can always start with a new branch and work on >>> that in parallel. >>> If there are enough people to work on it it should work. >>> >>> [1] - https://www.openhub.net/p/pax-web >>> [2] - https://www.openhub.net/p/pax-web/estimated_cost >>> [3] - http://ci.ops4j.org/jenkins/job/org.ops4j.pax.web/1028/testReport/ >>> >>> >>> >>> >>> 2016-10-02 15:39 GMT+02:00 iJava : >>> >>> Hi Achim Could you say (from the top of your head) approximatively how many hours may these changes need - 100/1000/5000/1? Best regards, воскресенье, 2 октября 2016 г., 15:40:23 UTC+3 пользователь Achim Nierbeck написал: > Sounds like a good and interesting idea ... > Right now only from the top of my head: > The Pax-Web Runtime and therefore the different Implementations aren't > made for this right now. So this would need a complete rewrite of how > we're > handling it. Another point would be how would web and white-board > extender > work with it. We could think about wiring those two closer to the core. > Never the less an application deploying servlets will always need to > add > the virtual host environment, working with defaults could take care of > that. > > We could consider to start this with a complete rewrite of Pax-Web and > therefore aim for a 7.0. > > BUT ... I fear I won't have enough time to takle this. Considering the > amount of time I spent in the past and about what it would take to > have all > the functionalities of Pax-Web re-written, and especially with my > $dayJob + > Family. > > regards, Achim > > > > 2016-10-02 5:35 GMT+02:00 Niclas Hedhman : > > Honestly, if this is to be fixed, I think Pax Web should support >> Managed >> Service Factory, and instantiate separate virtual host services >> according >> to a provided configuration. That configuration should contain which >> WAB(s) >> goes into that virtual host, together with any other virtual host >> configuration. >> >> To me, that seems to be the right solution forward, maintains OSGi >> compatibility, doesn't introduce new config args on WABs and doesn't >> treat >> "one domain" different than another. >> >> I think the tricky bit is to make the default case and the MSF >> instantiations play nicely with each other, but that is an design >> implementation detail at this stage. >> >> Cheers >> Niclas >> >> On Sat, Oct 1, 2016 at 4:49 PM, iJava wrote: >> >> I analyzed situation again and I am sure I am right. How I explain >>> this >>> - if *only* >>> web-contextpath is used then all
Re: PAX:WEB How many war bundles do pax-web support?
Here are my two cents Regarding the whiteboard-extender, I was actually thinking of moving this into the webcontainer, because due to the whiteboard-dto spec those two are closely related anyways. My idea was to deprecate the (upcoming) WhiteboardManager-service right away in order to merge those two modules in a 7.0 release. So that might solve one pain-point. But another question is: do we need to rewrite everything in order to get a feature which might no be needed? Without knowing the business-case behind registering multiple contexts with the same name in different virtual-hosts, I still think that there are much cheaper alternatives: everything today moves away from heavy-installations (AppServers) in favor of dedicated containers. With OSGi and Pax-Web you can easily spawn multiple VMs, and have some proxy/webserver in front which manages the site/domain to look like one. regards Marc Am Sonntag, 2. Oktober 2016 15:39:45 UTC+2 schrieb iJava: > > Hi Achim > > Could you say (from the top of your head) approximatively how many hours > may these changes need - 100/1000/5000/1? > > Best regards, > > воскресенье, 2 октября 2016 г., 15:40:23 UTC+3 пользователь Achim Nierbeck > написал: >> >> Sounds like a good and interesting idea ... >> Right now only from the top of my head: >> The Pax-Web Runtime and therefore the different Implementations aren't >> made for this right now. So this would need a complete rewrite of how we're >> handling it. Another point would be how would web and white-board extender >> work with it. We could think about wiring those two closer to the core. >> Never the less an application deploying servlets will always need to add >> the virtual host environment, working with defaults could take care of that. >> >> We could consider to start this with a complete rewrite of Pax-Web and >> therefore aim for a 7.0. >> >> BUT ... I fear I won't have enough time to takle this. Considering the >> amount of time I spent in the past and about what it would take to have all >> the functionalities of Pax-Web re-written, and especially with my $dayJob + >> Family. >> >> regards, Achim >> >> >> >> 2016-10-02 5:35 GMT+02:00 Niclas Hedhman: >> >>> Honestly, if this is to be fixed, I think Pax Web should support Managed >>> Service Factory, and instantiate separate virtual host services according >>> to a provided configuration. That configuration should contain which WAB(s) >>> goes into that virtual host, together with any other virtual host >>> configuration. >>> >>> To me, that seems to be the right solution forward, maintains OSGi >>> compatibility, doesn't introduce new config args on WABs and doesn't treat >>> "one domain" different than another. >>> >>> I think the tricky bit is to make the default case and the MSF >>> instantiations play nicely with each other, but that is an design >>> implementation detail at this stage. >>> >>> Cheers >>> Niclas >>> >>> On Sat, Oct 1, 2016 at 4:49 PM, iJava wrote: >>> I analyzed situation again and I am sure I am right. How I explain this - if *only* web-contextpath is used then all war bundles (wabs) are inside one domain. Obvious if you need more then one domains (virtualhosts) this limitation is unpleasant. So I am sure that when bundle is deployed it must have *two* settings: Layer one - virtualhosts (plural) Layer two - web-contextpath. In this case the deployer has all the advantages. He can create N sites And inside every virtualhost he can make N contexts if he needs. I am sure that this functionality must be developed. Pax-web is great product and with such functionality it will have all main functionality of a good web server. I would be glad to hear others opinion about such New Feature. Best regards, пятница, 30 сентября 2016 г., 18:14:33 UTC+3 пользователь iJava написал: > Ok Achim. > > I understood the situation. You know the architecture of pax-web well. > Could you say - how difficult > it can be to make some extender (plugin etc) to link wabs not to > web-contextpath but to virtualhosts > and to make them all work with one port like it is in usual web > servers (for example apache). > Please, note I don't care about specification - I care about normal > work. > > Best regards, > > пятница, 30 сентября 2016 г., 18:06:23 UTC+3 пользователь Achim > Nierbeck написал: >> >> I never said Pax-Web is a complete replacement for GlassFish, >> it's a WebContainer for OSGi environments, which fulfills the OSGi >> spec. >> It uses Jetty, Undertow or Tomcat to do so. AND it gives you most of >> the benefits of those underlying servers in the >> same way. If you're not satisfied because you expect something >> different. I'm
Re: PAX:WEB How many war bundles do pax-web support?
Ok, Achim. Thank you for detailed answers. I currently need and use only the following subprojects: pax-web-api-6.0.0-SNAPSHOT.jar pax-web-deployer-6.0.0-SNAPSHOT.jar pax-web-descriptor-6.0.0-SNAPSHOT.jar pax-web-extender-war-6.0.0-SNAPSHOT.jar pax-web-extender-whiteboard-6.0.0-SNAPSHOT.jar pax-web-jetty-6.0.0-SNAPSHOT.jar pax-web-jsp-6.0.0-SNAPSHOT.jar pax-web-runtime-6.0.0-SNAPSHOT.jar pax-web-spi-6.0.0-SNAPSHOT.jar Firstly I will take some time to study them. If after analysis I can find and suggest some solution which will require not so much time I will discuss it with you. But I say in advance that if I will do something, it will be linked only with subprojects I need and use. I think the reason can be easily understood. Best regards, On 02.10.2016 19:35, 'Achim Nierbeck' via OPS4J wrote: About a person year, with the prerequisite that the current implementation is well known. 1) Not really, certain people working on the code are paid to do so. 2) Nope, it's an open source project most people working on this do this in their private time. Sometimes those people (like me) don't even work on any related stuf anymore. 3) Hard to tell ... as we have people which get engaged for certain topics which move off after a certain amount of time Here's a complete list of recent developers [1] regards, Achim [1] - https://github.com/ops4j/org.ops4j.pax.web/graphs/contributors 2016-10-02 18:20 GMT+02:00 Pavel Kastornyy: Achim, thank you for the information. So about one person year if I understand you right. Could you also shortly answer the following questions: 1) is there any financial help from any companies? 2) has the community tried to draw investments into the product? 3) how many active developers are there at this time? Best regards, On 02.10.2016 18:47, 'Achim Nierbeck' via OPS4J wrote: wow, that's a tough one ... :D if you take a look at Openhub [1] ... it'll tell you it took about 57 years [2], or at least it's the amount of work worth it ;) Anyway it's hard to estimate as I've spent the last 6 years improving on it. Never the less if I would work on it full time 8h day with the knowledge I have right now. One person could redo it within maybe a year. But it's a wild guess ... might be faster, might take longer ... Of course I would expect it to have the same features and possibilities it has which makes it different to other implementations of the HttpService etc. Right now we have about 500 unit and integration tests running with every build [3]. That functionality I would expect to be available after the re-write ;) BUT, the nice thing is. We can always start with a new branch and work on that in parallel. If there are enough people to work on it it should work. [1] - https://www.openhub.net/p/pax-web [2] - https://www.openhub.net/p/pax-web/estimated_cost [3] - http://ci.ops4j.org/jenkins/job/org.ops4j.pax.web/1028/testReport/ 2016-10-02 15:39 GMT+02:00 iJava : Hi Achim Could you say (from the top of your head) approximatively how many hours may these changes need - 100/1000/5000/1? Best regards, воскресенье, 2 октября 2016 г., 15:40:23 UTC+3 пользователь Achim Nierbeck написал: Sounds like a good and interesting idea ... Right now only from the top of my head: The Pax-Web Runtime and therefore the different Implementations aren't made for this right now. So this would need a complete rewrite of how we're handling it. Another point would be how would web and white-board extender work with it. We could think about wiring those two closer to the core. Never the less an application deploying servlets will always need to add the virtual host environment, working with defaults could take care of that. We could consider to start this with a complete rewrite of Pax-Web and therefore aim for a 7.0. BUT ... I fear I won't have enough time to takle this. Considering the amount of time I spent in the past and about what it would take to have all the functionalities of Pax-Web re-written, and especially with my $dayJob + Family. regards, Achim 2016-10-02 5:35 GMT+02:00 Niclas Hedhman : Honestly, if this is to be fixed, I think Pax Web should support Managed Service Factory, and instantiate separate virtual host services according to a provided configuration. That configuration should contain which WAB(s) goes into that virtual host, together with any other virtual host configuration. To me, that seems to be the right solution forward, maintains OSGi compatibility, doesn't introduce new config args on WABs and doesn't treat "one domain" different than another. I think the tricky bit is to make the default case and the MSF instantiations play nicely with each other, but that is an design implementation detail at this stage. Cheers Niclas On Sat, Oct 1, 2016 at 4:49 PM, iJava wrote: I analyzed situation again and I am sure I am right. How I explain
Re: PAX:WEB How many war bundles do pax-web support?
About a person year, with the prerequisite that the current implementation is well known. 1) Not really, certain people working on the code are paid to do so. 2) Nope, it's an open source project most people working on this do this in their private time. Sometimes those people (like me) don't even work on any related stuf anymore. 3) Hard to tell ... as we have people which get engaged for certain topics which move off after a certain amount of time Here's a complete list of recent developers [1] regards, Achim [1] - https://github.com/ops4j/org.ops4j.pax.web/graphs/contributors 2016-10-02 18:20 GMT+02:00 Pavel Kastornyy: > Achim, thank you for the information. So about one person year > if I understand you right. > > Could you also shortly answer the following questions: > > 1) is there any financial help from any companies? > 2) has the community tried to draw investments into the product? > 3) how many active developers are there at this time? > > Best regards, > > > > On 02.10.2016 18:47, 'Achim Nierbeck' via OPS4J wrote: > >> wow, that's a tough one ... :D >> >> if you take a look at Openhub [1] ... it'll tell you it took about 57 >> years >> [2], or at least it's the amount of work worth it ;) >> >> Anyway it's hard to estimate as I've spent the last 6 years improving on >> it. Never the less if I would work on it full time 8h day >> with the knowledge I have right now. One person could redo it within maybe >> a year. But it's a wild guess ... might be faster, might take longer ... >> Of course I would expect it to have the same features and possibilities it >> has which makes it different to other implementations of the HttpService >> etc. Right now we have about 500 unit and integration tests running with >> every build [3]. >> That functionality I would expect to be available after the re-write ;) >> >> BUT, the nice thing is. We can always start with a new branch and work on >> that in parallel. >> If there are enough people to work on it it should work. >> >> [1] - https://www.openhub.net/p/pax-web >> [2] - https://www.openhub.net/p/pax-web/estimated_cost >> [3] - http://ci.ops4j.org/jenkins/job/org.ops4j.pax.web/1028/testReport/ >> >> >> >> >> 2016-10-02 15:39 GMT+02:00 iJava : >> >> Hi Achim >>> >>> Could you say (from the top of your head) approximatively how many hours >>> may these changes need - 100/1000/5000/1? >>> >>> Best regards, >>> >>> воскресенье, 2 октября 2016 г., 15:40:23 UTC+3 пользователь Achim >>> Nierbeck >>> написал: >>> Sounds like a good and interesting idea ... Right now only from the top of my head: The Pax-Web Runtime and therefore the different Implementations aren't made for this right now. So this would need a complete rewrite of how we're handling it. Another point would be how would web and white-board extender work with it. We could think about wiring those two closer to the core. Never the less an application deploying servlets will always need to add the virtual host environment, working with defaults could take care of that. We could consider to start this with a complete rewrite of Pax-Web and therefore aim for a 7.0. BUT ... I fear I won't have enough time to takle this. Considering the amount of time I spent in the past and about what it would take to have all the functionalities of Pax-Web re-written, and especially with my $dayJob + Family. regards, Achim 2016-10-02 5:35 GMT+02:00 Niclas Hedhman : Honestly, if this is to be fixed, I think Pax Web should support Managed > Service Factory, and instantiate separate virtual host services > according > to a provided configuration. That configuration should contain which > WAB(s) > goes into that virtual host, together with any other virtual host > configuration. > > To me, that seems to be the right solution forward, maintains OSGi > compatibility, doesn't introduce new config args on WABs and doesn't > treat > "one domain" different than another. > > I think the tricky bit is to make the default case and the MSF > instantiations play nicely with each other, but that is an design > implementation detail at this stage. > > Cheers > Niclas > > On Sat, Oct 1, 2016 at 4:49 PM, iJava wrote: > > I analyzed situation again and I am sure I am right. How I explain this >> - if *only* >> web-contextpath is used then all war bundles (wabs) are inside one >> domain. >> Obvious if you need more then one domains (virtualhosts) this >> limitation is >> unpleasant. So I am sure that when bundle is deployed it must have >> *two* >> >> settings: >>Layer one - virtualhosts (plural) >>Layer two - web-contextpath. >> In this case the
Re: PAX:WEB How many war bundles do pax-web support?
Achim, thank you for the information. So about one person year if I understand you right. Could you also shortly answer the following questions: 1) is there any financial help from any companies? 2) has the community tried to draw investments into the product? 3) how many active developers are there at this time? Best regards, On 02.10.2016 18:47, 'Achim Nierbeck' via OPS4J wrote: wow, that's a tough one ... :D if you take a look at Openhub [1] ... it'll tell you it took about 57 years [2], or at least it's the amount of work worth it ;) Anyway it's hard to estimate as I've spent the last 6 years improving on it. Never the less if I would work on it full time 8h day with the knowledge I have right now. One person could redo it within maybe a year. But it's a wild guess ... might be faster, might take longer ... Of course I would expect it to have the same features and possibilities it has which makes it different to other implementations of the HttpService etc. Right now we have about 500 unit and integration tests running with every build [3]. That functionality I would expect to be available after the re-write ;) BUT, the nice thing is. We can always start with a new branch and work on that in parallel. If there are enough people to work on it it should work. [1] - https://www.openhub.net/p/pax-web [2] - https://www.openhub.net/p/pax-web/estimated_cost [3] - http://ci.ops4j.org/jenkins/job/org.ops4j.pax.web/1028/testReport/ 2016-10-02 15:39 GMT+02:00 iJava: Hi Achim Could you say (from the top of your head) approximatively how many hours may these changes need - 100/1000/5000/1? Best regards, воскресенье, 2 октября 2016 г., 15:40:23 UTC+3 пользователь Achim Nierbeck написал: Sounds like a good and interesting idea ... Right now only from the top of my head: The Pax-Web Runtime and therefore the different Implementations aren't made for this right now. So this would need a complete rewrite of how we're handling it. Another point would be how would web and white-board extender work with it. We could think about wiring those two closer to the core. Never the less an application deploying servlets will always need to add the virtual host environment, working with defaults could take care of that. We could consider to start this with a complete rewrite of Pax-Web and therefore aim for a 7.0. BUT ... I fear I won't have enough time to takle this. Considering the amount of time I spent in the past and about what it would take to have all the functionalities of Pax-Web re-written, and especially with my $dayJob + Family. regards, Achim 2016-10-02 5:35 GMT+02:00 Niclas Hedhman : Honestly, if this is to be fixed, I think Pax Web should support Managed Service Factory, and instantiate separate virtual host services according to a provided configuration. That configuration should contain which WAB(s) goes into that virtual host, together with any other virtual host configuration. To me, that seems to be the right solution forward, maintains OSGi compatibility, doesn't introduce new config args on WABs and doesn't treat "one domain" different than another. I think the tricky bit is to make the default case and the MSF instantiations play nicely with each other, but that is an design implementation detail at this stage. Cheers Niclas On Sat, Oct 1, 2016 at 4:49 PM, iJava wrote: I analyzed situation again and I am sure I am right. How I explain this - if *only* web-contextpath is used then all war bundles (wabs) are inside one domain. Obvious if you need more then one domains (virtualhosts) this limitation is unpleasant. So I am sure that when bundle is deployed it must have *two* settings: Layer one - virtualhosts (plural) Layer two - web-contextpath. In this case the deployer has all the advantages. He can create N sites And inside every virtualhost he can make N contexts if he needs. I am sure that this functionality must be developed. Pax-web is great product and with such functionality it will have all main functionality of a good web server. I would be glad to hear others opinion about such New Feature. Best regards, пятница, 30 сентября 2016 г., 18:14:33 UTC+3 пользователь iJava написал: Ok Achim. I understood the situation. You know the architecture of pax-web well. Could you say - how difficult it can be to make some extender (plugin etc) to link wabs not to web-contextpath but to virtualhosts and to make them all work with one port like it is in usual web servers (for example apache). Please, note I don't care about specification - I care about normal work. Best regards, пятница, 30 сентября 2016 г., 18:06:23 UTC+3 пользователь Achim Nierbeck написал: I never said Pax-Web is a complete replacement for GlassFish, it's a WebContainer for OSGi environments, which fulfills the OSGi spec. It uses Jetty, Undertow or Tomcat to do so. AND it gives you most of the benefits of those underlying
Re: PAX:WEB How many war bundles do pax-web support?
wow, that's a tough one ... :D if you take a look at Openhub [1] ... it'll tell you it took about 57 years [2], or at least it's the amount of work worth it ;) Anyway it's hard to estimate as I've spent the last 6 years improving on it. Never the less if I would work on it full time 8h day with the knowledge I have right now. One person could redo it within maybe a year. But it's a wild guess ... might be faster, might take longer ... Of course I would expect it to have the same features and possibilities it has which makes it different to other implementations of the HttpService etc. Right now we have about 500 unit and integration tests running with every build [3]. That functionality I would expect to be available after the re-write ;) BUT, the nice thing is. We can always start with a new branch and work on that in parallel. If there are enough people to work on it it should work. [1] - https://www.openhub.net/p/pax-web [2] - https://www.openhub.net/p/pax-web/estimated_cost [3] - http://ci.ops4j.org/jenkins/job/org.ops4j.pax.web/1028/testReport/ 2016-10-02 15:39 GMT+02:00 iJava: > Hi Achim > > Could you say (from the top of your head) approximatively how many hours > may these changes need - 100/1000/5000/1? > > Best regards, > > воскресенье, 2 октября 2016 г., 15:40:23 UTC+3 пользователь Achim Nierbeck > написал: >> >> Sounds like a good and interesting idea ... >> Right now only from the top of my head: >> The Pax-Web Runtime and therefore the different Implementations aren't >> made for this right now. So this would need a complete rewrite of how we're >> handling it. Another point would be how would web and white-board extender >> work with it. We could think about wiring those two closer to the core. >> Never the less an application deploying servlets will always need to add >> the virtual host environment, working with defaults could take care of that. >> >> We could consider to start this with a complete rewrite of Pax-Web and >> therefore aim for a 7.0. >> >> BUT ... I fear I won't have enough time to takle this. Considering the >> amount of time I spent in the past and about what it would take to have all >> the functionalities of Pax-Web re-written, and especially with my $dayJob + >> Family. >> >> regards, Achim >> >> >> >> 2016-10-02 5:35 GMT+02:00 Niclas Hedhman : >> >>> Honestly, if this is to be fixed, I think Pax Web should support Managed >>> Service Factory, and instantiate separate virtual host services according >>> to a provided configuration. That configuration should contain which WAB(s) >>> goes into that virtual host, together with any other virtual host >>> configuration. >>> >>> To me, that seems to be the right solution forward, maintains OSGi >>> compatibility, doesn't introduce new config args on WABs and doesn't treat >>> "one domain" different than another. >>> >>> I think the tricky bit is to make the default case and the MSF >>> instantiations play nicely with each other, but that is an design >>> implementation detail at this stage. >>> >>> Cheers >>> Niclas >>> >>> On Sat, Oct 1, 2016 at 4:49 PM, iJava wrote: >>> I analyzed situation again and I am sure I am right. How I explain this - if *only* web-contextpath is used then all war bundles (wabs) are inside one domain. Obvious if you need more then one domains (virtualhosts) this limitation is unpleasant. So I am sure that when bundle is deployed it must have *two* settings: Layer one - virtualhosts (plural) Layer two - web-contextpath. In this case the deployer has all the advantages. He can create N sites And inside every virtualhost he can make N contexts if he needs. I am sure that this functionality must be developed. Pax-web is great product and with such functionality it will have all main functionality of a good web server. I would be glad to hear others opinion about such New Feature. Best regards, пятница, 30 сентября 2016 г., 18:14:33 UTC+3 пользователь iJava написал: > Ok Achim. > > I understood the situation. You know the architecture of pax-web well. > Could you say - how difficult > it can be to make some extender (plugin etc) to link wabs not to > web-contextpath but to virtualhosts > and to make them all work with one port like it is in usual web > servers (for example apache). > Please, note I don't care about specification - I care about normal > work. > > Best regards, > > пятница, 30 сентября 2016 г., 18:06:23 UTC+3 пользователь Achim > Nierbeck написал: >> >> I never said Pax-Web is a complete replacement for GlassFish, >> it's a WebContainer for OSGi environments, which fulfills the OSGi >> spec. >> It uses Jetty, Undertow or Tomcat to do so. AND it gives you most of >> the benefits of those