This was our third implementation using eZ publish and was done at a time when the
product was rapidly evolving and there was lack of important documentation. Very few
details existed on how to best implement the system and we had to work out things along
the way both in terms of process and implementation. We learned some valuable lessons
that made future implementations much easier.
The most important lesson we learned was the value of receiving all of the content
upfront. The requirements contained some complex rules for publishing content in
different areas. This meant the specification became more complex, in particular the
content model and display templates, which also made development harder. Having more
of the content upfront, we could have better tested the validity of the requirements and
simplified some of the rules, which would have made the specification and development
phases easier without affecting the quality of the end result.
Another lesson learned was with our approach to building the HTML prototype. We used
traditional layout techniques, i.e. heavy use of tables. We have since found that using an
XHTML/CSS approach (i.e. all layout and styling defined in the CSS) we can
dramatically reduce the amount of time it takes to convert the HTML prototype to eZ
publish templates. It also reduces the page size, and increases accessibility and
performance. This method fits in well with the modular approach to content that eZ
publish dictates and has been very successful in subsequent projects. We are now
converting the CFD site to XHTML/CSS to make any further changes easier to manage.
On the whole, we've found that the more time spent upfront, analyzing requirements,
understanding the client, their needs, and their content leads to a much smoother
implementation and a better end result. The better we understand the client's domain and
their content, the better the solution we can come up with. In subsequent projects,
although the time spent in the requirements and specification phases has slightly
increased, the time taken in the development phases has decreased significantly.
Finally, the task of creating, entering, and shaping content should not be underestimated.
No matter how well structured the information architecture, how efficient the
implementation, how elegant the design, or impressive the use of XHTML/CSS, it all
becomes meaningless without quality content. The result of all this hard work is merely a
system for managing content—and the system is only as good as the content itself.