Re: cvs commit: jakarta-tomcat/src/share/org/apache/tomcat/core ServletWrapper.java

["Craig R. McClanahan" <Cr...@eng.sun.com>]

costin@locus.apache.org wrote:

> costin      00/04/17 14:28:56
>
>   Modified:    src/share/org/apache/tomcat/core Tag: TOMCAT_31
>                         ServletWrapper.java
>   Log:
>   Fix for destroy() - note that even with this fix, reloading is _not_
>   production ready and is _not_ a "supported" feature of tomcat 3.1.
>

I am -0 (leaning towards -1) on including any feature we don't think has been tested enough to rely on.

I am *very* -1 on leaving a feature like this in, without warning users what to expect.  Please add a note to the "KNOWN BUGS AND ISSUES"
section of the release notes document ("src/doc/readme") that tells users what they should expect on this issue, or anything else where you
have concerns.

In the absence of any documentation to the contrary, users will assume all features are "production ready".  And last time I checked, very
few users read the CVS commit logs :-).

Craig McClanahan

Re: cvs commit: jakarta-tomcat/src/share/org/apache/tomcat/core ServletWrapper.java

["Craig R. McClanahan" <Cr...@eng.sun.com>]

Costin Manolache wrote:

> "Craig R. McClanahan" wrote:
>
> > costin@locus.apache.org wrote:
> >
> > > costin      00/04/17 14:28:56
> > >
> > >   Modified:    src/share/org/apache/tomcat/core Tag: TOMCAT_31
> > >                         ServletWrapper.java
> > >   Log:
> > >   Fix for destroy() - note that even with this fix, reloading is _not_
> > >   production ready and is _not_ a "supported" feature of tomcat 3.1.
> > >
> >
> > I am -0 (leaning towards -1) on including any feature we don't think has been tested enough to rely on.
>
> That's the way open source works - I don't understand how a feature is
> supposed to be tested if we don't include it ?
>

No argument on adding it ... in between releases.  When you come up to a release, it's time to make "it's in
or it's out" decisions.  If the decision is "it's in" (the usual case), we as developers should be pretty
confident about the quality of what we've posted, or we should help users set their expectations
appropriately.

>
> Right now the only thing that is tested is compatibility with the watchdog and
> tomcat test suite

And the fact that these suites didn't catch a lot of the recent bugs points to incompleteness in the test
suites, but that's another issue :-)

> - everything else is experimental, and will be experimental
> till enough people will review the code, test it in strange configurations
> and enough time will pass. Probably around tomcat 3.3 we might be
> able to claim that we had good testing on those features...
>

I wish I could believe that, but it seems more likely that there will be a bunch of new features that will be
the focus of attention.

>
> It works enough to let people enhance it - and that's the most important
> thing.
>

Successful open source projects, including our namesake one (Apache), as well as our predecessor (Apache
JServ), draw a distinction between a "release" and other interim builds (nightly, milestones, betas, and
release candidates).  The assumption that users want to make is that the entire set of "stuff" in this
release has been tested together, and is therefore stable enough to rely on.  That's why we do a "feature
freeze" when working up towards a release -- focus shifts for a short while on fixing what's already there,
instead of adding or changing more things.

If we didn't do this, there would be no reason to distinguish the release that Sam is going to cut tomorrow
(3.1) from a nightly build that just happened to be created on the same day -- and users would be correct in
assuming they could never count on Tomcat's stability, because some part of it's code would *always* be in
flux.

>
> Costin
>

Craig McClanahan

Re: cvs commit: jakarta-tomcat/src/share/org/apache/tomcat/core ServletWrapper.java

[Costin Manolache <co...@costin.dnt.ro>]

"Craig R. McClanahan" wrote:

> costin@locus.apache.org wrote:
>
> > costin      00/04/17 14:28:56
> >
> >   Modified:    src/share/org/apache/tomcat/core Tag: TOMCAT_31
> >                         ServletWrapper.java
> >   Log:
> >   Fix for destroy() - note that even with this fix, reloading is _not_
> >   production ready and is _not_ a "supported" feature of tomcat 3.1.
> >
>
> I am -0 (leaning towards -1) on including any feature we don't think has been tested enough to rely on.

That's the way open source works - I don't understand how a feature is
supposed to be tested if we don't include it ?

Right now the only thing that is tested is compatibility with the watchdog and
tomcat test suite - everything else is experimental, and will be experimental
till enough people will review the code, test it in strange configurations
and enough time will pass. Probably around tomcat 3.3 we might be
able to claim that we had good testing on those features...

It works enough to let people enhance it - and that's the most important
thing.

Costin