Re: [xml-dev] Designing an experiment to gather evidence on approaches to designing web services

Richard Salz scripsit:

> A large financial services firm in the Boston area converted from a 
> variety of data silo's to an all-XML storage and data-interchange 
> infrastructure and was able to decommission a significant number of 
> servers.
> 
> Is that evidence or anecdote?  How can you tell?  Suppose I removed all 
> the vagueness and replaced it with concrete numbers?

Anecdote either way.  Scientific (as opposed to legal) evidence requires a
planned test set between two or more alternatives, holding everything not
being tested as equal as possible, with enough individual tests to make
statistical inference meaningful, and with safeguards to avoid deceit or
(more difficult) self-deceit by the investigator.

In the nature of things, such gold-standard evidence testing is not
always possible.

-- 
So they play that [tune] on                     John Cowan
their fascist banjos, eh?                       
        --Great-Souled Sam                      http://www.ccil.org/~cowan