On 03/03/2016 16:41, Mark Thomas wrote: > On 03/03/2016 15:41, Christopher Schultz wrote: >> Mark, >> >> On 3/1/16 5:12 PM, Mark Thomas wrote: >>> To summarise where I think this discussion is going: >>> >>> - Create 8.5.x from 9.0.x with the following changes >>> - revert all changes to spec APIs >> >> I would argue that anything that has been added (in TC9) can stay; only >> revert the removals and possibly any deprecations. > > There are no removals. > > The spec APIs have to be correct. Our spec API JARs are used in various > places and if we starting shipping a Servlet 3.1+ API we'll create the > potential to cause all sorts of problems. > >>> - make any necessary changes to work with Java 7 >>> >>> - Release 8.0.x and 8.5.x in parallel for ~6 months then stop 8.0.x >>> releases >> >> Does this have implications for whether or not we can claim >> spec-compatibility? Tomcat 8.5 would be >> servlet-3.1-except-for-that-Java-7-support-thing... so can we claim that >> Tomcat 8.5 officially supports the servlet 3.1 specification? Do we even >> care? > > 8.5 will be as spec compliant as 8.0. > - It will run on Java 7. > - The spec APIs will be identical between versions. > - It will, as far as we know, be compliant with all > four specs it implements > > We haven't had access to the TCK for quite some time. It is a useful > additional check that in an ideal world I'd like to have back but not > having it doesn't appear to be causing the project any harm.
One thought I had for BIO support was that we could add something like this to handle the case where the user has explicitly selected BIO public class Http11Protocol extends Http11NioProtocol { public Http11Protocol() { super(); log.warn("BIO connector removed. Using NIO instead."); } } And the same for AJP. Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org