Stephen McConnell wrote: >> Just so long as there was one real banana. One Tomcat that always >> worked the best, always was the default, etc. Just like the current >> Tomcat ! >> >> So you can have your cake and eat it too. Just have a default >> container, one with everything, that does everything, one that >> follows all the rules. Lesser or special purpose containers wouldn't >> hurt a bit. You might want to hide them from the main pages a bit, >> but that is a different matter. > > I'm totally with you here - except for one point. If you rephrased the > above so that the default container is the reference implementation > (RI), set's the standard for any container - but "does not do > everything" - then I think we have something better than Tomcat. We > have the ability to add the container functionality when and where we > need and if we can assume that variant containers are based on a > strict container validation test-suite - then the user does not care > about the container. > > Assume that anything that passes the test-suite must run in a kernel. > If you look at this with the pricipal of kernel and container > separation - then the user end up saying - ok - here is the kernel, > here is some container (I don't care which container because it no > longer matters) and I drop the container onto the kernel and it runs - > guaranteed - every time.
Agreed enthusiastically, but only after the default container is a done deal. I used the shorthand *everything*, and you properly pointed out that there are some features that would best be left out of any default anything, and instead placed in the laboratory versions. What you are talking about here is above and beyond, and much better, agreed. But you have to crawl before you can run. Let's get to a crawl, then a walk, first. -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
