[microformats-discuss] DUHPER (WAS: XHTML-REST, STRUM, DETH, etc.)

[Dr. Ernie Prabhakar]

Hi Christopher,

On Oct 18, 2005, at 11:02 AM, Christopher St John wrote:

> On the other hand, the way state-transfer mechanisms (in the general
> sense, which includes RPC and REST) are generally used assumes that  
> the data is properly
> structured. The issue is that since the information is designed to  
> be used without
> human intervention, small mistakes can cause bad, hard to detect  
> failures. I'm not
> saying the mechanisms _must_ be used that way, it's just what  
> people normally
> expect. It isn't the existing usage pattern for state tranfer, RPC- 
> ish or REST-ian,
> either one.

You make an excellent point, and one we need to walk very carefully  
with.   The key question, IMHO, is whether "properly structured ==  
rigidly structured."   Today, (almost) everybody seems to assume you  
need a well-formed and valid XML document in order to ensure data  
integrity.   But is that true? Is it even possible to get a non- 
trivial community to agree on a single version of a schema, and then  
not change it incompatible ways?

The microformat (DUHPER? XOXR?) way is to define protocols tightly,  
but allow individuals the freedom to be more or less rigid, as  
necessary.   For example, I might say that my calendaring services  
requires six different vevent fields, and reject any submissions that  
contain less. That is my right!  However, if you want to send a few  
extra fields, I should just ignore them rather than complaining about  
an invalid document.   Conversely, if you want to create a service  
that only requires two fields (summary & dtstart), feel free.

The point, as I think has been made elsewhere before, is that it is  
really useful to be able to separate ''mechanism'' from ''policy.''  
By providing a sufficiently generic mechanism (CSS classes + XHTML  
tags), the hope is that individuals and micro-communities can adopt  
their own policies ad hoc on top of that without introducing needless  
incompatibility.

Will it work? I don't know, but I honestly can't think of anything  
else that even has a chance...

-- Ernie P.


------------
Ernest N. Prabhakar, Ph.D. <drernie at opendarwin.org>
Ex-Physicist, Marketing Weenie, and Dilettante Hacker
Probe-Hacker blog: http://www.opendarwin.org/~drernie/