Lists Home |
Date Index |
Mike Champion wrote:
> That's definitely something to think about carefully. My original
> point was more in reference to the argument that disallowing relative
> URIs in namespaces would somehow undermine the "Architecture of the
> Web." I haven't read Roy Fielding's work, but I doubt if it says
> much about XML namespaces, and I'm almost positive that no one with
> Fielding's clarity of vision is guiding the development of the XML
> family of specs.
Remember that the big corporations didn't care much about HTTP so it's
designers were allowed to work in peace. Not unlike the early days of
> Let's examine the design principles of "The Web":
> http://www.w3.org/DesignIssues/Principles.html is worth reading
> frequently and memorizing the key points:
> Keep it Simple
> Design in Modularity
> Be Tolerant in What You Accept, Strict in What You Produce
> Do One Thing Well
> Favor Simplicity Over Power
Well these aren't really what I'm talking about when I talk about Web
design. I'm talking more about these:
> The Web has a hard core -- HTTP and URIs definitely, (X)HTML and XML
> 1.0 to a very great extent, and all sorts of other things such as
> namespaces and schemas and xlink that fuzz out from there. I would
> question whether those peripheral parts of the web that do not
> obviously conform to its design principles should be considered to be
> "designed properly" until a lot more experience says so.
> Until that happens, I'm very leery of arguments along the lines of "I
> understand the design of The Web, and doing XYZ violates it." Newton,
> as I understand it, invented much of his mathematics to support his
> alchemical research ... The math stood the test of time, and the
> alchemy is clearly nonsense, even though Newton presumably thought
> they were integral parts of an overall "architecture."
True enough! I am precisely proposing to go back and study the math --
the parts that work astoundingly well -- so that our newer work is more
like calculus than alchemy. It isn't the case that understanding the
design of the Web gives one perfect insight into what should or should
not be added, any more than understanding the design of a single
software system lets you add extensions without ever making a mistake.
Nevertheless, you cannot hope to come up with a coherent system
*without* understanding the design.
And the issues are not always so abstract as relative URIs in
namespaces. Recently there was an issue with a sort of abuse of HTTP in
XForms. That was a web architecture issue which I believe will be
resolved to everybody's satisfaction. Now that I think about it, the
whole argument could probably have been short-circuited with a reference
That's the benefit of architectural documents.