"Geir Magnusson Jr." wrote:
> Jon Stevens wrote:
> > on 4/5/01 7:34 AM, "Paulo Gaspar" <[EMAIL PROTECTED]> wrote:
> > > 1- I am a bit skeptic about pure pull;
> >
> > Why?
Note that the skeptic qualms of Paulo is stated to apply to pure pull,
whereas below I'm proposing a coded/selectable preporocessing with
an attached pure pull view.
>
> In some ways, it breaks MVC. With a broad brush :
>
> For example, the canonical Model 1 JSP example seems to be
> authentication - in a Model 1 JSP app, each page has to do the same
> authentication work. Error prone and brittle.
>
> With pull, if you are going to just dump the kitchen sink into the
> context, then each piece has to deal with the same problem.
>
> With non-pull, the Controller in conjunction with the model can make
> decisions about what goes into the context. Sure, some parts would
> still be user sensitive, but you can head of a lot of it at the top, and
> localize, rather than spread across all things put into the context.
This is why in my TemplateServlet (or ApplicationServlet) I use an
init.vm template at the beginning, then a preprocessing request.vm
template which does the common work (assisted by an XML configuration).
A response template takes over the view rendering and finally a
layout template puts it into an HTML body.
See the sequence diagram depicting this at:
ftp://ftp.dfd.dlr.de/put/reck/VTL_MVC_seq_diagr.gif
This architecture make the above quoted arguments against the pull
model obsolete. Same applies to the turbine architecture where
a java coded screen (with an optional catch-all Default.java) takes
care of the preprocessing, with additional security and action
modules.
Long live the pull model!
Anyone interested in parts of this software should please contact me
directly.
:) Christoph
