Re: draft findings on Unsafe Methods (whenToUseGet-7)

At 09:45 AM 4/18/02 -0700, Paul Prescod wrote:
> > The section immediately preceding the one you cite mentions "actions they
> > might take which may have an unexpected significance to themselves or
> > others" -- in a web context, it seems to me that any action which changes
> > what is visible on the web is potentially significant, hence my attempt to
> > characterize it in terms of visible side effects.
>
>What if following a link on the Web had the semantic "inject patient
>with drug". That may not have visible side effects but is still a
>serious, dangerous side effect. Conversely, incrementing a page hit
>counter is not dangerous but IS web-visible.

OK, I accept both points.

But, in assessing an application (e.g. does it conform with architectural 
principles), what tests can I apply?  I think that web-visibility is a 
reasonable first approximation.

In the case of a counter, if it's a hit-counter then it's "mostly 
harmless".  But if the number incremented is the price for a service, 
that's less so.  The point is that I can't think of a technical way to 
distinguish between the two.

And to complicate the web-counter case, contemplate this scenario:  "Acme 
corporation offers a prize of $1000 to the millionth visitor to its web 
site"...  (Just to illustrate that the benign side effect of changing a 
counter might be important in some circumstances.)

SO what am I trying to say?  I'm wary of the value of technical design 
principles that cannot be technically assessed.  My entry to this debate 
was trying to be clear about what was the meaning of "safe", where I saw 
two possibilities that could apply.

#g


-------------------
Graham Klyne
<GK@NineByNine.org>

Received on Thursday, 18 April 2002 15:03:28 UTC