Everything on this list will eventually get done. If you need a particular set of functionality, I'm available, my rates are reasonable,
and you will be enshrined in the documentation as a Really Cool Person or Organization.
I've had a number of people ask when I can port the wftk to native Java. I'd love to, but I have a family to feed. I would like
to propose a wftk-j consortium to help me with the design and maybe all together chip in enough money to keep me off the streets for
a month or so. If you're interested, tell me. If I get a couple of takers, I'll create a wftk-j
consortium page with everybody's names, a timeline, progress reports, and so forth. I think J2EE would benefit from a solid free
workflow engine. Note: even if everyone is completely silent on this
topic, I will be starting on wftk-j in July September 2002. Or whenever.
As noted elsewhere, the wftk consists (or will consist) of two central
modules. The initial v1.0 release includes only one of those, the core engine. The second is the
repository manager, and in July 2002 I will first release the repmgr-v1.0 properly, then I will
begin work on wftk-v1.5, which will be the integration of the two systems. The repository manager
is slick because it represents the other end of what I think of as the "action/data" dichotomy --
while the wftk core organizes action rather nicely, it doesn't do a good job organizing data and
datasources. The repository manager addresses that weakness.
There are several examples of workflow definition languages out there. Yes, I do hope to address every last one of them eventually.
The WfMC (Workflow Management Coalition) is an industry standards group which has defined five interfaces for interacting with
workflow engines. There is as yet no reference implementation. It would be cool if wftk could be that implementation.
The interfaces are: (1) procdef definition language, (2) workflow client app API, (3) application invocation, (4) interfaces
to other workflow engines, and (5) admin and monitoring. Not a one of these is particularly horrible to implement, and I'd love
to do it. If you're interested in seeing progress on this front, please leave a suitcase with a large number of unmarked small
bills under the bridge. Tell no-one, and don't involve law enforcement.
IBM has recently been making noises on the workflow front, and has come up with some XML definition languages which look
attractive: WSFL (Web Services Flow Language), for instance, would be nice to implement.
(Business Process Modelling Language) -- same story, different people. I'd like to do something good here, too. In general,
note that getting wftk to work with a different procdef language is essentially a matter of writing a new pdrep adaptor, one
which compiles the language in question and arrives at the wftk's procdef language to present to the core engine. Simple.
By "scripting wrapper" I mean an API wrapper for a scripting language. The Python wrapper, for instance, works well already. The
Tcl wrapper works under AOLserver and is still not 100% stable, it seems at the moment.
Perl: use Workflow;
C++ classes (and documentation of classes already designed)
Visual Basic/ActiveX object
I used to do a lot of VB work, and I've just recently had a couple of inquiries about using wftk from VB. Good idea! I'm
now working on it.
Somehow, although I had even started doing the research on a PHP module, I forgot it on this
page. But now I have a motivator for PHP! So this will be moving forwards in priority.
One thing obviously and egregiously missing from the wftk at present is any form of a UI beyond the command-line tools.
A WebDAV wrapper (based on the Medusa server) has been proposed as a dandy way to make information visible from
the repository manager. WebDAV is really neat. Consider the ability to save a file from (say) Photoshop or Word,
at which point a workflow process is automatically started. How cool is that? I love this job.
PHP might actually turn out to be the first UI implementation, actually. Didn't expect that.
Popup window UI
The popup UI is a lightweight C++/wxWindows app which fronts for the repository manager (and thus indirectly for the wftk).
The Python wrapper is reaching maturity, which means Zope should be easy to do.
Medusa rocks. Thanks to Jo Meder, I'm nearly finished with a Medusa-based repmgr server.
The Tcl/AOLserver wrapper is also reaching maturity and is in use at one production facility for data maintenance. This
application uses the repmgr wrapper.
With or without a .NET wrapper, an IIS UI would be a useful thing.
This requires the C++ classes above.
Making CVS workflow-aware is something I've been thinking about for a while but just
haven't had the time to deal with; CVS is not the friendliest piece of software to work with.
(Of course, once it's all set up it just blends into the background, which is why nobody
spends much time changing it -- you don't even remember it's there.) The idea here is that
I could grant person X an update privilege level which, instead of being simply "update",
would take each update set that the user performed, call it all a single patch set, and then
create an approval process for the patch set based on that user's group (i.e. different
users could have different approval processes.) Until approval succeeds, that user would see
their own changes in the repository, but nobody else would (this would require some sort of
module finagling or something; I don't know yet). Once approval was granted (possibly of a
changed patch set) then everybody else would see those changes upon checkout. Note
that the submitting user needs to get a set of updates if their patch set gets changed during
approval. Finally, if the patch set is flat-out rejected, then the user needs to get a set
of changes to return the code to its pre-changed state. A real can of worms, but boy would
it ever be convenient!
I don't think I need to belabor this point. See the examples list. All those systems should
be built as demo systems.
There are plenty of places to extend functionality without sweeping changes in the wftk (that's the whole point). And there
are some reorganization tasks to be done. Some of the tasks I list here won't make a lot of sense unless you're pretty familiar
with the code.
Rebuild dsrep and pdrep adaptors on the list adaptor.
Move Oracle database functionality to list adaptor class from dsrep class.
MAPI notification adaptor.
And more interestingly, wouldn't it be great to set up wftk as a trigger handler for PostgreSQL? A workflow-aware database.
That would be so ... words fail me.
OpenLDAP for perms, users adaptors.
gdbm for file-system-based indexing.
... (add your favorite here) ...
There is no shortage of peripheral systems with which wftk should know how to interact.