General talk about software patterns

[Arno Haase <Arno.Haase AT haase-consulting.com>]

Moises,

I agree that organizing prior art and knowledge, especially patterns, is 
an important and relevant issue. One approach that is explored at the 
pattern conferences is the organization of patterns into pattern 
languages, refactoring existing patterns based on what is learned on the 
way, and exploring sequences in which patterns are actually applied in 
real scenarios. This is being done for several technical domains that 
are relevant for architecture.

But I do not see how a hierarchical (tree) structure of topics could 
hope to capture such a big and complex topic as architecture, esepcially 
aiming at having a single system for *all* kinds of systems. In my 
experience, there are some technical domains that are relevant for very 
many different systems (security, memory and general resource 
management, remoting, ...) but with different trade-offs - embedded 
systems have some very different forces from e.g. web shops. And then 
there is also the issue of interaction between process and architecture.

The hierarchical approach you suggest may work for very high-level 
descriptions of architectures, but I expect there to be redundancy / 
copy & paste when the level of detail goes beyond powerpoint architectures.

How do you propose to avoid this redundancy problem that arises when you 
 reach a level of detail when most of the architectural issues become 
cross-cutting? And, on a more general note, what is the primary 
dimension you would suggest for the specialization hierarchy in 
architectures you suggest?

Best regards

- Arno


Moisÿffffe9s Dÿffffedaz wrote:
> Hi all,
>
> On my opinion, at this moment, Software Architectures
> are so abstract that relatively it is not very useful
> for developers. All that a lot of people can read
> about this topic are generalities that are very far
> from day-to-day practitioners and software architects
> that work in software industries.
>
> Nevertheless, I think that we need architecture for
> understanding systems, organising its development, 
> and making reuse more real. But perhaps we need an
> intermediate (and new) concept to bring software
> architecture to software development. Describing
> different software architectures is not enough.
>
> I think that this new concept must to extend design
> patterns’ success. 
> This concept (that I have named ‘Archetypes’) is a
> form of documenting (in one document):
> Functional requirements.
> Logic architectures (layers, pipelines, etc).
> Physic architecture (real components, frameworks,
> deployment configurations, etc).
>
> The idea is develop a tree of documents related with
> areas of software (web portals, enterprise information
> systems). The final documents represent specific type
> of software system, with defined logic architecture
> and patterns, and with specific physic architecture
> (frameworks, components, etc). Higher documents in the
> tree will be only related to specific types of
> software systems, and logic architectures. 
>
> I developed this concept in a small article (It is not
> updated with the keyword ‘archetype’, I use (badly)
> the term ‘pattern language’):
> http://www.moisesdaniel.com/wri/htdsalp.html . An
> example of archetype would be an article that I wrote:
> ‘Enterprise Information Systems’ Architecture’
> http://www.moisesdaniel.com/wri/eisarc.html.
>
> I think that identifying main architectures and
> patterns is only a part of the real objective: we need
> methodologies that can manage in a direct way the
> architectural subject and I think that Archetypes can
> be a bridge between Software Architectures and
> Methodologies.
>
> What I’m searching at this moment is opinions about if
> this concept would be interesting and useful. 
>
> Excuse me if all what I’m commenting is not very
> reasonable.
>
> Best Regard,
> Moisés D. Díaz
>
>
>
>                 
> ______________________________________________
> Renovamos el Correo Yahoo!: ¡100 MB GRATIS!
> Nuevos servicios, más seguridad
> http://correo.yahoo.es
>
> _______________________________________________
> patterns-discussion mailing list
> patterns-discussion AT cs.uiuc.edu
> http://mail.cs.uiuc.edu/mailman/listinfo/patterns-discussion