perl-unicode

Re: In-Band Information Considered Harmful

1998-10-26 14:18:08
Tim Bray wrote :
|| At 10:03 AM 10/26/98 -0800, Stephen Zander wrote:
|| >Thus to make Perl REx effective, all of the datastream must be
|| >available to the programmer, because only the programmer knows what is
|| >or isn't "essential".  
|| 
|| Exact-a-mundo. -T.

It sounds to me like it is worse than that.  If "all of the
datastream" is not sufficient for the programmer to decide which
markup elements are significant (for the particular editted/analysis
operation he is trying to program), the best he can do is to work
correctly when *all* of the markup used is from a pre-defined set of
types that he knows how to handle and blindly hope that anything else
that comes along is irrelevant and ignore it.

Possibly, there can be some sort of a deal made, whereby any
additional markup defined by an author must be described in terms
that does allow the code to determine whether that markup is
significant (and if so, how to handle it).  That, however, supposes
that the author knows about the particular class of operations that
the programmer is doing and is will to assist.  As soon as a
programming task requires any additional dimension of interpretation
applied that hasn't been planned for by the author, the programmer
has to deal with extensions himself.  (The programmer basically has
to deal with a new spec every time an author adds something new.
That's worse than having just Microsoft and Netscape trying to extend
your problem space as fast as they can.)

-- 
objects:                                    | John Macdonald
    Think of them as data with an attitude. |   jmm(_at_)elegant(_dot_)com