Adam,

I looked through some of my archived emails on this subject.  A common
theme amongst the MyFaces developers (in addition to myself) was that
the ADF code should be refactored to use the MyFaces "commons" code
when possible.  At one point in time the ADF team seemed to agree with
this but reminded us that their may be cases where the ADF way was
superior to the MyFaces way and the refactoring would be done to the
existing MyFaces code in those cases.  (IMO a perfectly reasonable
point.)

At one point the ADF developers began to review the MyFaces code and
reported back that such a merger was going to be more difficult then
Oracle had originally thought.  Over time Oracle's position seemed to
have drifted away from these assuarances that the code would be
refactored (I found a private PMC email thread discussing this
changing position.)

I think there is a pretty strong interest in taking a look at the ADF
code at this point.  The only person to look at it from the MyFaces
PMC was myself and that was (as you described) the reverse engineering
of one of your more complicated components.

If Oracle's position is that they are not interested in significant
changes to their code to fit nicely with the rest of MyFaces, then my
personal position is that I am not in favor of ADF becoming part of
the MyFaces project.  I think its a major mistake to have 60% of
tomahawk components use one approach to solve a problem and 40% use a
completely different approach.  This will become a maintenance
disaster.

I'm also not wild about duplicate components.  So I think some
consolidation is in order.  Instead of two data tables, take the best
of both and have a single data table.  More is *not* better.  Not when
it comes to maintaining code and documentation.  I am still not wild
about tree and tree2.  We certainly don't need dataTable and
dataTable2, etc.

There are two important challenges here.

1.) Oracle needs to be ready, willing and able to do significant
refactoring of their code (beyond renaming the packages.)

2.) MyFaces committers (or potential committers) need to be ready to
step up and assist in the massive integration effort.  Significant
changes to the existing code will have to be made to accomodate the
new components and to take advantage of areas where the ADF approach
is superior.

So far Oracle seems to be backing away from #1 and we have one
volunteer for #2.  I don't even think its worth moving the code to the
incubator if we can't get more support in these two areas.

Please don't get the impression that I'm against ADF becoming part of
MyFaces.  Tthe idea is a great one in theory.  Oracle has some cool
components and a lot of JSF know how.  Oracle has been at the
forefront pushing JSF and MyFaces can only get better with people like
yourself involved in the project.

A lot of this has already been said on a private email chain with
Oracle way back when.  I think these issues need to be revisted again
- this time in public.  Not only am I concerned but the incubator and
project committees are getting jumpy.  What does the ADF team envision
happening here?

Regards,

Sean


On 12/29/05, Adam Winer <[EMAIL PROTECTED]> wrote:
> On 12/29/05, Sean Schofield <[EMAIL PROTECTED]> wrote:
> > > I seem to recall hearing that ADF uses a custom viewhandler, much like
> > > tobago.   That's going to make it difficult to trivially merge
> > > Tomahawk and ADF .
> >
> > You are absolutely correct.  This is a serious concern that has yet to
> > be addressed.  Oracle has said they would refactor to use the MyFaces
> > codebase but I have seen no evidence that this has (or will) take
> > place.
>
> Sean, I don't remember anyone making such a promise.
>
> What I know was said (when you started diving in to the
> implementation of the table component, I'm guessing after
> decompiling some of our source) was that we had a large base
> of rendering code built as an adapter layer around an old framework
> (UIX), and that this was going to be rewritten to go directly to
> JSF APIs.   A large amount of that has happened, incl. the table.
>
> (BTW, w/o having looked at the Tobago code - our custom
> ViewHandler is an adapter only;  we're still using ordinary JSPs
> - and Facelets - for our rendering.  I don't see that specifically
> as being of great concern.)
>
> -- Adam
>

Reply via email to