> From: Hans Bergsten [mailto:[EMAIL PROTECTED]]
>
> One of the most important things I'm trying to get a grip on is what
> the pros and cons are for the "community" (both developers and users)
> if we release the current HEAD as the next 3.x version, as opposed to
> continue to do bugfixes based on 3.2.1. From what I've seen
> so far, all
> the work you and others have done since 3.2 seems to be for
> the better;
> more structured, more comments, performance improvements,
> etc. But it's
> a *lot* of changes, and from my limited testing it's clear that there
> are new bugs introduced (no surprise, that's what happens when you do
> major refactoring). This means that we need active committers that
> understand the new architecture if we are going to make it the next
> 3.x. You can try to downplay your role in this as much as you
> like; the
> fact remains that 90% (or more) of all commits on this version have
> your name on them. So when you now say that you give up,
> you're making
> it much harder for me to consider a 3.x based on the HEAD
> code. Chances
> are that if you leave, the three or four other committers
> that have been
> actively working on the refactoring with you will go with
> you. In that
> scenario it's probably better for the community to continue the 3.x
> branch based on 3.2, since it's been battle tested and therefore is
> likely to have fewer bugs and it's compatible with the modifications
> and additions users have done privately.
Having worked with both 3.2 and 3.3 code, I don't consider the 3.2 as
being very maintainable. By this I mean, that it's code is still
convoluted enough to make bug fixing difficult if not near impossible
in some situations. Implementing bug fixes in 3.2.x will carry a higher
risk of introducing new bugs than in a released 3.3.
I know Craig ran into this when fixing BugRat Report #316. This was prior
to the initial Tomcat 3.2 release, where he appropriately added
"response.flushbuffer()" to RequestDispatcherImpl.include(). This fixes
the bug but adds a "quirk" to Tomcat's RD.include(). Given the choice of
trying to fix something like this in 3.2.x, or attempting a fix in 3.3,
I think I would have much better success in 3.3.
I think in the long run, the community will be better served by a released
3.3. It may have some different bugs, but I think it will eventually have
fewer bugs and quirks and be more maintainable as well.
The need here at SAS Institute with respect to the Java IDE I work on is
for a Servlet 2.2/JSP 1.1 container. I don't expect that to change for
at least 6 months to a year. As a result, my primary focus will remain
on Tomcat 3.x and don't plan on "leaving".
Cheers,
Larry
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, email: [EMAIL PROTECTED]
