On Fri, Aug 21, 2009 at 12:20 AM, Andreas Veithen <andreas.veit...@gmail.com > wrote:
> All, > > Just to clarify some things: > > 1) I explained the purpose of the fix in AXIS2-4465. I believe that > the description gives enough information about the why and the how of > the change. If not, point to the parts that are not clear. Please also > have a look at the linked issues to see the kind of problems the code > had before the fix. While doing the changes I inadvertently removed a > public method. As explained above, this can be easily fixed, and I > will do so. > > 2) I think there is no need to start yelling around about "merrily > chang[ing] APIs", "things are going to become a big mess" and "engage > the community before making such drastic changes". I did lots of fixes > in Axiom and Axis2 since Synapse trunk switched from Axis2 SNAPSHOT to > 1.5, and when Hiranya switched back to SNAPSHOT there were no > particular issues related to them, except for the issue discussed > here. I personally take the suggestion that my changes make of Axis2 a > big mess as an offense. If that is what some people think, I can stop > immediately to work on Axis2 and Axiom. With respect to engaging the > community, as mentioned above, the description of AXIS2-4465 provides > enough information about the reasons for the change and as Dims > pointed out, we have a commit-and-review policy. Since the change is > focused, not massive, can easily be reverted and contains an > appropriate level of Javadoc, I think that the community has > everything that is needed to review it. > > Finally, I could ask in reply how all these principles applied to the > change in r744900 [2], which incidentally is the second change that > caused problems when switching Axis2 versions on the Synapse trunk and > which did not address any bug at all. I may be wrong but I don't think Axis2 cluster management is an extension point at all. Synapse is the only project which directly uses this code for its load balancing stuff. And when the change at [2] was made even Synapse was not depending on Axis2 trunk and hence nothing really broke up. And since Azeez is a Synapse committer I'm sure when he made the above changes, he was planning to make the necessary changes in the Synapse trunk once we change the Axis2 version of Synapse to SNAPSHOT (and he actually did help me out a lot in making the necessary code changes in Synapse when I did the migration yesterday - See my svn commit log message at [3]). Also looking at the changes that we needed to perform on the Synapse trunk during migration, they were all rather small and insignificant changes like renaming of classes being referred, renaming a few methods being called etc. It was pretty easy to get Synapse trunk to compile with Axis2 snapshot. No changes were required to the design of any of the classes in Synapse. No new code were required to be written. Sorry for replying to something that was not meant for me. But since I was involved in changing the Axis2 version in Synapse trunk I thought I needed to speak up. Thanks, Hiranya [3] - http://svn.apache.org/viewvc?view=rev&revision=806071 > > > 3) The discussion in AXIS2-4465 pointed to the problem that even by > preserving the public API of AxisServlet, we will probably not be able > to avoid breaking subclasses. This is caused by the fact that > AxisServlet doesn't have an extensible design. If somebody has a > brilliant idea how to get around this problem (other than reverting > the fix and reopening the 6 issues it is supposed to solve), please > speak out. > > Andreas > > [1] http://markmail.org/message/62dixjx3qrqry3yr > [2] http://svn.apache.org/viewvc?view=rev&revision=744900 > > On Thu, Aug 20, 2009 at 18:08, Senaka Fernando<sen...@wso2.com> wrote: > > Hi Andreas, > > > > Just wondering what you are trying to achieve here. Is this related to > auto > > detection of ports as Hiranya pointed out? While I appreciate the effort > > you've put into doing something worthwhile, I believe that getting rid of > a > > public method in a class is not the correct thing to do. I believe that > what > > you have done here is the addition of a new portion of code. Can we make > the > > new portion of code optional? And leave the existing logic as it was? > Also, > > are you planning further changes to this class? if so, it would perhaps > be > > better to figure out a more elaborate solution, which safeguards both the > > existing level of extensibility of this class and also its public API. > > > > Thanks, > > Senaka > > > > On Thu, Aug 20, 2009 at 6:14 PM, Hiranya Jayathilaka < > hiranya...@gmail.com> > > wrote: > >> > >> Hi Andreas, > >> > >> By looking at the code I got the impression that HTTP transport > receivers > >> should extend the AxisServletListener class for your logic of port auto > >> detection to work. Is that correct? What happens if the transport > receivers > >> used do not extend this class? All request handler methods call the > >> preprocessRequest method which in turns run port auto detection. If the > >> transport receivers do not extend AxisServlerListener how is that > handled? > >> > >> Thanks, > >> Hiranya > >> > >> > >> On Thu, Aug 20, 2009 at 6:05 PM, Andreas Veithen > >> <andreas.veit...@gmail.com> wrote: > >>> > >>> Afkham, > >>> > >>> The only change I see in the public APIs is the disappearance of the > >>> initContextRoot method. We can easily fix this be restoring the > >>> original initContextRoot method and let the preprocessRequest method > >>> call initContextRoot. Do you see any other things to change? > >>> > >>> Andreas > >>> > >>> On Thu, Aug 20, 2009 at 13:45, Afkham Azeez<afk...@gmail.com> wrote: > >>> > Yes Dims. However, if everybody continues to merrily change APIs, > >>> > making public methods private & so on, things are going to become a > >>> > big mess. Axis2 provides public APIs, and those may be having > >>> > problems, but still they are public APIs. This is why you have to be > >>> > very careful when defining APIs; if you get them wrong, you may have > >>> > to live with it for a long time. > >>> > > >>> > Azeez > >>> > > >>> > On Thu, Aug 20, 2009 at 11:38 AM, Davanum Srinivas<dava...@gmail.com > > > >>> > wrote: > >>> >> Azeez, > >>> >> > >>> >> We are still following, commit-then-review right? > >>> >> > >>> >> thanks, > >>> >> dims > >>> >> > >>> >> On 08/20/2009 07:33 AM, Afkham Azeez wrote: > >>> >>> > >>> >>> Hi Andreas, > >>> >>> The changes you've done to the APIs as per > >>> >>> https://issues.apache.org/jira/browse/AXIS2-4465 badly breaks some > of > >>> >>> the projects that depend on Axis2. Please revert this, and please > >>> >>> engage the community before making such drastic changes in the > >>> >>> future. > >>> >>> > >>> >> > >>> > > >>> > > >>> > > >>> > -- > >>> > Thanks > >>> > Afkham Azeez > >>> > > >>> > Blog: http://afkham.org > >>> > Developer Portal: http://www.wso2.org > >>> > WSAS Blog: http://wso2wsas.blogspot.com > >>> > Company: http://wso2.com > >>> > GPG Fingerprint: 643F C2AF EB78 F886 40C9 B2A2 4AE2 C887 665E 0760 > >>> > > >> > >> > >> > >> -- > >> Hiranya Jayathilaka > >> Software Engineer; > >> WSO2 Inc.; http://wso2.org > >> E-mail: hira...@wso2.com; Mobile: +94 77 633 3491 > >> Blog: http://techfeast-hiranya.blogspot.com > > > > > -- Hiranya Jayathilaka Software Engineer; WSO2 Inc.; http://wso2.org E-mail: hira...@wso2.com; Mobile: +94 77 633 3491 Blog: http://techfeast-hiranya.blogspot.com
