On 9/27/06, Remy Maucherat <[EMAIL PROTECTED]> wrote:
Peter Rossbach wrote: > Hi Filip, > > I thing many people use the current cluster implementation. For those > people I want spend my time to sync two branches :-) > Some people ask me to start experiments with new Servlet and JSP API's > or the new connector implementation but with > the old cluster API. Let the user the choice,like APR or Java HTTP > Connector. This does not make much sense here, as the new implementation does the exact same thing as the other one (unlike the connectors). Did you test the new clustering ? How does it compare to the old one ? There's no need for two clustering implementations, so one of them needs to be chosen. The other one may be developed in the sandbox and provided as an alternative (in a separate download), but I think that's time which would be wasted. BTW, a major release is also the best opportunity to make some changes in configuration. If it's not done now, then we would have to keep both for a long time.
IMO both clustering implementations should be provided as separate downloads. No need to put them in sandbox ( since they were released in the past, it's 'proven' code ). Not sure if they should be in separate repos or it's ok to have both of them in the main tree. Costin
