W3C home > Mailing lists > Public > www-tag@w3.org > March 2006

RE: Rule of Least Power

From: Bullard, Claude L \(Len\) <len.bullard@intergraph.com>
Date: Wed, 8 Mar 2006 08:43:10 -0600
Message-ID: <15725CF6AFE2F34DB8A5B4770B7334EE0BB1FF3E@hq1.ingr-corp.com>
To: "Harry Halpin" <hhalpin@ibiblio.org>
Cc: <www-tag@w3.org>, <RogerCutler@chevron.com>

That's dismissive, Harry.  If you have 
never passed an object as a named argument, 
or haven't worked with transforms (eg, XSLT), 
you won't understand up and down translation 
or object-oriented communications.  Given 
a very large scale, contexts are contracts, 
thus, DTDs, Schemas, etc., which are in fact, 
powerful means to manage complexity.

Simple is better when building for scale  
because testing is cheaper than contract 
negotiations and contexts vary. 

That is the RLP in a nutshell.  

Reuse comes from stripping intentions at the cost of 
weakening the contract because process is expensive 
to enforce across boundaries/edges of 
systems.  The web is a system with weak edges 
because the network contracts are deliberately 
and smartly separated from the semantic 
and pragmatic contracts.

Ok.  This is a case where the names on the 
finding carry more weight than the actual 
concepts and fields.  That has a tendancy 
over time to erode the authority of the 
body and the credibility of the authors. 
Caveat emptor.  

len


From: Harry Halpin [mailto:hhalpin@ibiblio.org]

While the stream-of-consciousness examples and objections from Len are
interesting, it should be pretty self-evident by now that the simplest
architectures/programming languages/mark-up languages that does the job
tend to do better on the Web than more complex ones, and insofar as the
Rule of Least Power states that fact, I think many people would agree. 
Also, by happy chance, we also tend to be able to be able to
mathematically prove more about simple things than more complicated
things, re previous discussions over Turing-completeness, functional
programming, and so on. If you disagree, that's fine. However, for a more
productive discussion it would be useful to point out either concrete
Web-based use cases where the Rule is clearly violated, and
then point how amendments could be made that deal with such cases. In
this case matter it would be productive to hear
what cases were brought up by folks at Chevron, either on or off-list.

                             -harry

Bullard, Claude L (Len) wrote:

>That's precisely backwards, Harry.  Passing an object passes 
>intent (functions) with the data.  That is as expressive as 
>it gets if you want a computer to respond with better 
>reliability and ontological commitment.
>
>To communicate as much as possible, one would send the 
>object, not just the data.  Organic systems have to 
>work hard to understand each other.  Computer systems 
>don't of necessity but do because of human application. 
>Well-designed BI systems get this and experienced designers 
>are careful about the granularity of data collected over 
>the reaction time given noisy data.  (In simple terms: 
>if you overreact, you fly it into the runway.  If you 
>underreact, it flies itself into the runway.  See applications 
>of real time systems to the Airbus disasters.)
>
>HTML is an SGML application language (there is no such 
>thing as an SGML dialect unless you count the XML subset). 
>HTML isn't a programming language.  Neither is the parent. 
>ISO was scrupulous about that.    Had HTML been the most 
>expressive means, we wouldn't be using XML for on the 
>wire messaging and HTML for a lossy downtranslation target. 
>You are confusing ubiquity of the presentation context 
>with the ubiquity of the transmission context.  SGML won.
>
>Maximum reuse comes of stripping intentions in the 
>expression so that it can be interpreted in the context 
>of the consumer more readily.   Least power means greatest 
>number of possible interpretations.  It puts the semantic 
>back into superposition requiring measurement.  This principle 
>means Heisenbergian observers where the context of observation 
>determines the value with the probability of distortion 
>some contexts preclude others.
>
>In a command and control system, that can be a very bad idea 
>or a very good one but this comes down to context management, 
>not a principle of least power. 
>
>So in the sense of ubiquity, this is a weak principle and should 
>be marked as such informatively. 
>
>len
>
>From: www-tag-request@w3.org [mailto:www-tag-request@w3.org]On Behalf Of
>Harry Halpin
>
>At least from my reading of the Rule of Least Power (and it took me a
>while to finally understand what was being said), I think
>the point is fundamentally on open world systems like the Web to
>communicate as much as possible about your process (be that programming
>language, web page design language, or whatever) upfront. It's a bit
>different that Grice's maxim, since Maxim #2 is "Do not make your
>contribution to the conversation more informative than necessary" - and
>on the Web you never know how much might be necessary, so it's far
>better to communicate as much as possible, and so restrain your choices
>of languages to those that have properties that are known in advance.
>And maybe I'm twisting it a bit, but I think this is a good argument
>against using Java applets and for using REST architecture to deliver
>information. But it could also mean mean using XHTML instead of some
>more expressive SGML dialect for your web-page. Overall, the principle
>seems fairly sensible and historically has been proven right again and
>again.
>
>-harry
>Bullard, Claude L (Len) wrote:
>
>  
>
>>The problem with that example is it contrives HTML to 
>>be a programming language.   It isn't.
>>
>>Simple examples that do not obtain rob a principle 
>>of its surface credibility. The fact that a principle 
>>is easily misapplied may mean it is inadequately 
>>formulated or its application context isn't common. 
>>That means it begs its reason to be included among 
>>general principles.
>>
>>As far as I can tell, this isn't a principle of 
>>the computer science of building web applications, 
>>but of politeness in a communicative context.  Again, 
>>Grice's Maxims are equally applicable and better argued.
>>
>>len
>>
>>
>>From: www-tag-request@w3.org [mailto:www-tag-request@w3.org]On Behalf
>>    
>>
>Of
>  
>
>>noah_mendelsohn@us.ibm.com
>>
>>I think it's easy to make the case that such pages would, at least in
>>those 
>>senses, be less valuable than pages with similar content conveyed in
>>HTML. 
>>
>>
>>This is a somewhat contrived example, in that few of us are tempted to 
>>build static pages entirely in JavaScript, but that's the point.  In
>>many 
>>cases, the Rule of Least Power brings you to solutions that have 
>>commonsense value.  I think that if one looks to such obvious examples
>>one 
>>may convince people that they are already acting on the Rule of Least 
>>Power.  If so, maybe they can start to understand how the Rule might
>>guide 
>>them toward useful decisions that they might not have made as well
>>without 
>>it. 
>>
>> 
>>
>>    
>>
Received on Wednesday, 8 March 2006 14:43:26 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:39 GMT