I have worked to assemble web based technology for many years and I have some thoughts on the mess that has become web design. Initially there are segments of the technologies and skills that separate the work force along with perceptions. Analogies in this area are useless due to assumed self competence in the technology by all parties. I have been the first on several technologies and laughed at more than my share of head hunters. Watching industry leading e-tailers fail with new deployments and working to correct the issue while enabling the people is what I have done.

Design
Just being a part of the total project design often gets too much attention as it is the hands on interface for the hire level decision makers. Some industries with incorrect assumptions about their audience the industry standards omit design. Lucky for us the lack of design does clean the disrespecting groups from the commercial playing field. The Internet and its users punish sales for any reason and often for none at all.
Technology
Many of the technologies used in a website or larger eCommerce project are well established technologies dating back to nearly a century. UML with it's descendant SGML are simple markup languages that in some cases predate my parents birth. These are the ideas that form modern XML, HTML and CSS. The code exist and must exist. There is no presence without the tech unlike design and information. Development languages for the web could be and should be broken into segments to enable confidence in a project. When the target audience is using an aging renderer do not target the platform, enable the platform after the standards are supported.
Information
Absolutely vital information is often omitted from the data that is supplied to the designers and technologist. The most innovative projects I have seen involve dialog with designers and technologist for the creation of the project scope. I often say "it has to work before you can do xyz" which entails functional over features.
Testing
While many humors and cruel things could be said for the data collection of results at the end of a project we will take a positive approach. The user votes with a mouse and ignores most of our hard work. We need multiple functional ways to preform every action along path ways that transverse the process or movement though a site. Success is reached when testing exceeds the effort of the other aspects of design, technologies, and information management.
http://lathama.com
~