The default is one hour (much less than MySQL's eight hour max idle time) but
it can be configured with `max_idle_minutes` (supplied along with the other
connection provider configuration properties).
The powers that be decided that it would be cool to break everything in the
entire Java universe because they didn't want to negotiate with Oracle the
permission to take over the javax.servlet namespace.
So now we all get to either have two versions of all of our web-related
libraries, or just pick a time to break every downstream dependency and force
them to jump through this renaming hoop as well. Today is the day that we're
forcing the hoop jumping.
h/t to fourbites for starting this whole ball rolling.
Otherwise if samskivert.jar is loaded via the root classloader, this will fail
to find classes and resources in the servlet/webapp classloader. Let's not
require that samskivert.jar be duplicated into the webapp in cases where we're
doing complex things like hosting webapps inside larger app servers. That can
cause other fiddly problems with mismatched duplicate copies of classes.
- Nix attempts to link to ancient, no longer in existence, Oracle hosted
Javadocs.
- Manually include code to attach javadocs and source jars during release,
because we no longer get that from the OSS parent pom. We can't use the OSS
parent pom because it redirects release stuff to the ossrh Maven Central
servers that no longer exist. Would be nice if they made and published a new
OSS parent pom. Shrug.
This is the preferred way to obtain a JDBC connection in this millenium. No
longer does one manually instantiate the JDBC driver by name. The JDBC driver
jar file will have a mapping that indicates that it handles a particular type
of JDBC URL and the JDBC infrastructure will handle instantiating the driver.
This makes them invisible to JavaScript (which we don't use) to avoid the
danger of someone injecting JavaScript somehow and stealing auth cookies. It's
a dangerous world out there.
- De-static checkedApply(). It should be overrideable and know
other things about the ObserverList that it belongs to.
- Created an overrideable observerForLog().
- Have the DerefOp include the original Op's toString() in its own.
The SiteResourceLoader is some ancient stuff that's not used anywhere any more,
and there's no use in keeping this crufty crap around to test it. I'd nix it
too but I don't want to cause yet more trouble.
TestUtil was weird and a bad idea and good riddance to it as well.
Michael Bayne