Re[2]: Is "simplicity" a useful architectural constraint?

On Thursday, January 03, 2002, 3:53:46 AM, Mike wrote:

>> From: Jeff Lowery [] 
MD> [...]
>> Maybe a scale of grokability is the way to go... hell, if I 
>> understand it, it's got to be simple. If only Roy T. Fielding 
>> understands it... well, back to the drawing board. 
MD> I agree - if only one or two implementors understand how to properly
MD> interact with their system via a protocol/interface/whatever then back to
MD> the drawing board. If many many implementors can understand how to interact
MD> via a protocol/interface/whatever, then it has a chance.

This is where the implementation reports at advancement to Last Call,
CR and PR, and the Candidate Recommendation phase in the W3C process
shows its worth.

A useful metric, not required in the process (nor should it be, but
its still useful) is to do interop studies between implementations
from member organisations of the working group, on the one hand, and
as implementations from the public (who have only the spec to go on,
not the shared experience of hundreds of telcons and tens of face to
face meetings and whiteboard chats). Its a good way to reveal the
"well I guess we should say that explicitly, then" gaps.

Another useful metric of grokability (do we really have to eat them?)
is the speed with which translations into other languages occur.
Obviously this is affected by length and by developer interest, and
random chance, too. But a translator, unlike an implementor, has to
tackle an entire spec not just the parts that solve their particular
problem. Translators make excelemt proof readers (in the sense of
'what does this really say' not spelling etc).


Received on Friday, 4 January 2002 07:39:33 UTC