[
Lists Home |
Date Index |
Thread Index
]
At 16:31 08/12/2002 -0500, Andrzej Jan Taramina wrote:
>But not every operation needs to be a separate process in a pipeline
>(though it
>could be). In the example of validation versus schemas, it might be easier to
>have one validation process, but have different parameters used to configure
>that process (validate structure only, validate structure AND types..etc.).
>
>Modularity has a way of solving such thorny issues, and pipelines are just
>another form of modularity.
>
>What is interesting is that this approach is finding favour with the Aspect-
>Oriented Programming crowd, where method calls and aspects (interceptors)
>are, in a sense, pipelined together dynamically. The architecture of the
>JBoss
>project is an excellent example of such thinking and provides many benefits
>and flexibilities.
I think it is useful to think of them as separate stages. I see it as a compile
time optimization to coalesce multiple stages into a single stage without
intermediate serialization.
Many moons ago - before XML and indeed before SGML, I used this
from assembler using a technique called Jackson Inversion. Works
a treat.
I presented a paper at an SGML conference once called What have Michael
Jackson and SGML got it common. A lot more than many people
think:-)
Sean
|
