[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

RE: Too much power? was RE: 2007 Predictions

• From: "Michael Champion" <mc@x...>
• To: "'XML Developers List'" <xml-dev@l...>
• Date: Mon, 22 Jan 2007 09:00:17 -0800

> -----Original Message-----
> From: noah_mendelsohn@u... [mailto:noah_mendelsohn@u...]
> Sent: Monday, January 22, 2007 7:44 AM
> To: Kurt Cagle
> Cc: Michael Champion; XML Developers List
> Subject: Re:  Too much power? was RE:  2007
> Predictions
 
> The Rule does not discourage using either Turing-complete languages in
> general or imperative languages in particular where they are the best
> means of capturing behavior, some algorithm.  What it does encourage is
> the transmission of data and information (and logic or algorithms when
> practical) in simple declarative forms when those simple forms are good
> enough.  On the Web, it's a good thing if that data is given a URI that
> can facilitate access to and resuse of that information (in this case,
> your flight lists.)

As I noted in another post, I came to that same conclusion after reading the
final draft of the finding. Sorry for going off on a rant.  Still, I would
have been a lot less confused / annoyed if it had been called something like
the "Principle of Declarative Data addressed with a URI".  (I can't remember
why the TAG changed "Principle" to "Rule" in the title).  "The Rule of Least
Power" sounds a lot more awesome than the actual advice being offered really
is.

Also, I wish this kind of architectural advice was expressed in a contingent
way, e.g.   

- If  information is not highly confidential, and potentially reusable by
multiple parties or usefully bookmarkable by a single party, consider making
addressable via a URI

- Consider transmitting information in simple declarative form when that is
"good enough", e.g. the information can be easily represented in linear or
tabular form using standards such as HTML or XML+CSS.

- Consider transmitting information as XML + an XSLT stylesheet to render it
if a) it can be assumed that an XSLT processor that supports the stylesheet
is installed on the client and b) offloading the display computation to the
client is a good tradeoff across client side processing power, server side
processing capacity, and network bandwidth.

- If on the other hand, information is intended for close user interaction
or cannot be displayed using widely deployed software, consider embedding it
in some sort of safe executable such as a script language or applet. 
 
And so on.

 So many of the religious wars over URIs, SOAP endpoints, binary MXL, XLINK,
and so on that the TAG has gotten involved in seem to me based on misguided
hopes for uniformity / universality by some of the parties to the
discussion. Clearly stating the conditions under which some principle is
assumed to hold can pre-empt a lot of contention.  If the advice of the TAG
or some other group formed to offer authoritative statements on some
contentious matter sounds like the ethereal pronouncements of architecture
astronauts, it will probably be ignored.  If on the other hand, it sounds
like the concrete wisdom of experience, it will probably be followed.   

• Follow-Ups:
  • RE: Too much power? was RE: 2007 Predictions
    • From: noah_mendelsohn@u...

• References:
  • Re: Too much power? was RE: 2007 Predictions
    • From: "Kurt Cagle" <kurt.cagle@g...>
  • Re: Too much power? was RE: 2007 Predictions
    • From: noah_mendelsohn@u...