Hi folks,

If you look at the issue at hand, its not about just maintaining two
p2-profiles. Event though it sounds simple enough, its get more complicated
after branching. After any component is branched relevant features should
be branched and p2-profiles update. If its a cost cutting component then we
need to modify 13 product plus the services which is not easy at all. If
you recall during the last 3.2.x release towards the end we had two
versions of a single jar packed in the services and ultimately Shankar had
to write a script to ensure only that the latest jars were packed, right in
the middle of the release. Though we expect p2 to handle these stuff, it
not that straight forward when it comes to releases.

Besides its not only about two p2-profiles. There are bin.xml + the distro
to manage as well. If a new config file is introduced we need to update 26
bin.xml files! Its hard enough having to update 13 in the first
place. There are plenty of occasion where I find either a config file or db
scripts are missing or out of date in the services. When not in a release
rush Stratos team can sort those out, but its hard for the Stratos team to
keep in tab with 13 services all the time.

Also there is the issue of integration test. As it stands at the moment,
the services are not even started when running integration test, leave
alone testing multitenant aspects. We've been down the path where we expect
the service to just work if the product is working, but we also know
through experience that its not the case. One reason for that is the
services are not up to date plus the services has its own complexities
(causing bundle wiring issues and what not!). The worst part is we only
begin to uncover these issue after the product release is over and about to
do a stratos release. Then again nobody can expect teams to look into the
services when the product release is round the corner.

Of course I am not trying to make any excuses nor find fault. The obvious
issue here is having to maintain two distros vs one. I fully understand
that there are complexities in merging the product and the service, but
IMHO it would be beneficial in the long run. If we postpone the merging
effort to say six months, its going to be even harder and not easier, so
its better to get it done sooner if possible. And if we do managed to pull
it off, then its one less release cycle!

At the end of the day, if you log into the service as super tenant that
basically should serve as the relevant product. Bottom line is If ST
functions doesn't work as in the product, the the relevant service is
broken !! food for thought :) ..
_______________________________________________
Dev mailing list
[email protected]
http://wso2.org/cgi-bin/mailman/listinfo/dev

Reply via email to