Remy Maucherat wrote: >> For example if we fix something in jk - should we have to make a full >> new release of tomcat ? Same for jasper, catalina, etc. > > Yes, we do. We release stable builds based on multiple components. We > can't support pick and choosing (latest big example: Xerces, which you > couldn't upgrade to the latest release). That's a guarantee to users > that we have tested that particular combination. How we handle the thing > internally is irrelevant, we have to present users with a single archive > containing everything needed.
It's not about "pick and choose" - we control the list that makes up a stable release. I think we're looking at this problem from very different angles. I won't use xerces as an example ( it's a trap :-), but commons-logging If some bugs are fixed (that affect tomcat) - we can recommend people to update commons-logging to the specific version that we tested and includes the fixes. If we find a bug in jasper - we can test the fix against the "stable" release and make a mini-release with only jasper. At any givent time, "Tomcat stable" will be the latest major release ( 4.1.24 ) plus any additional patches that we test. All OS vendors have "patches" that include updates to individual components ( well, Microsoft has the huge service packs, but most other just update very specific components ). Most of our components are relatively independent of each other and we have reasonably stable APIs, so "pick and choose on your own" could work, but we can't support it. >> As long as we test the new jar against the stable version - and _we_ >> recommend the components that should be mixed - I think it is far >> better way to upgrade. > > This is too complex, and not practical. We already have problems with > regressions in stable releases, people not testing anything, etc, even > with the current system. What you propose would just make things worse. > > I don't like your ideas today ;-) Wait until tomorrow to submit them :) :-) Costin --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
