[PLUG] UML users?

Tue May 20 13:22:02 UTC 2003

With regard to UML and version control, it is indeed a problem, but not
a show stopper. The greatest benefit with UML is to create a model for a
developer team and verify it against the requirements, which involves
people outside the development team. In this scenario the model is the
more and less fixed frame for the subsequent implementation. IOW,
changes are kept to a minimum and documented because they affect the
team and possibly require communication with outsiders (i.e. when
business rules or a RDBMS structure is affected). 

In some cases, i.e. a Java enterprise application where business objects
and rules are implemented as beans (classes) in a distributed
environment UML is an integral part of the development with the finished
model representing a major development effort and component. 

Personally I also use it for smaller one-man projects, because it saves
me from typing code.

Lastly, a diff is possible on the XMI file that an UML tool should be
able to create. This XML file contains the entire model and is text.

Hope that didn't upset anyone now...

Holger

On Tue, 2003-05-20 at 10:34, Phil Tomson wrote:
    On Tue, 20 May 2003, Jeme A Brelin wrote:

    >
    > On Mon, 19 May 2003, Phil Tomson wrote:
    > > No, I'm not trying to trash it.  I'm trying to find out why people use it.
    > >
    > > For me it seems like a step in the wrong direction, but if I can
    > > understand why people choose to use it maybe I can understand why people
    > > want to go that way and then offer them alternatives ;-)  I really do
    > > think there might be better methodologies....
    >
    > I think Rich's point is valid despite the "shove off" language.
    >
    > You say that it seems like a step in the wrong direction before really
    > knowing much at all about it.  You assert counter-points to the individual
    > statements of those who use the system but you fail to take those
    > statements in aggregate appreciate the way the tool is used regadless of
    > how you might use it yourself.  You can't learn the usefulness of
    > something if you're just trying to find out how useless it is.

    Oh, my!  Email is such an imprecise communication medium, eh?  It seems
    I've come across as some sort of very closed minded person.

    Actually, I've read some of the articles on UML by Booch (several years
    back, I admit).  Recently we had some discussion about UML on another
    list I'm on and we had a bit of a debate - I guess there were a couple
    of other old hardware engineers on that list because I found I had some
    agreement about not wanting schematics for software.

    I've actually considered learning UML, but time is limited so I need to be
    choosy about what I invest in.  That's why I try to find out what is
    motivating people to move in that direction.  I'm coming from a different
    industry where I witnessed the reverse happen: there was a movement away
    from graphical design descriptions toward textual ones, and at the time
    this was a rather radical shift.  I witnessed this shift first-hand
    during my career.  When I graduated in the mid-80s we drew schematics
    exclusively - very few people had even heard of HDLs or any kind of
    textual description of circuits.  By the mid-90s HDLs were making inroads
    and mixed schematic (graphical) and HDL (text) designs were being done.
    Now in the ASIC and FPGA world it's very rare to find schematics being
    used at all - the paradigm has completely shifted and hardware engineers
    are definitely more productive because of it.  There is no way that the
    types of chips being designed now could be done in a timely fashion using
    schematics.

    So please don't think of my criticism as merely taking potshots at
    something I know nothing about; I'm merely offering my observations based
    on what I've seen happen in the world of chip design over that last
    fifteen years.  You can take those observations as instructive as related
    to software design, or you can write them off as totally unrelated, that's
    up to you.

    >
    > Now, I have never personally worked with UML and so I really have no
    > comment on its usefulness as tool for the programmer.  However, I have
    > seen PLENTY of bad code badly documented and I welcome anything that will
    > help describe the design of a program for future maintainers and
    > modifiers.  I think this is really a difference in philosophy more than a
    > difference in technology.  To me, the social aspects of how a thing is
    > done far supercede the efficiency or expedience of how a thing is done.
    > I personally believe that the purpose of work is to make the world a more
    > pleasant place.  The inconvenience of using a cumbersome tool (if, indeed,
    > that's what we've got here) in the design phase seems to be a way of
    > paying a small cost in pain up front for alleviating the long pains of
    > future developers who must work from your code.  Bad code never dies and a
    > little inconvenience today could prevent a lot of inconvenience in all of
    > our tomorrows.
    >

    Schematics were fine for documenting small circuits.  If you only needed
    two or three pages of schematics it was a reasonable way to understand how
    the circuit worked.  But for large systems it was extremely awkward.  I
    recall that in my first job out of school they handed us a huge three-ring
    binder full of hundreds of pages of schematics for a very complex chip
    tester and we were supposed to go off and study them.  I'd take them home
    and read them trying to figure the thing out.  You'd be chasing one signal
    and then it would get to the end of the page and there would be a little
    note about the other pages the signal went to... oh, the pain, I'm trying
    to forget it. ;-)

    Anyway, please excuse what seemed to be a viceral reaction against UML; it
    just seems as though we hardware engineers have been there and done that
    and I'm not really sure I want to go back there.

    Phil

    _______________________________________________
    PLUG mailing list
    PLUG at lists.pdxlinux.org
http://lists.pdxlinux.org/mailman/listinfo/plug