Re: PIFLE (was Re: Are PIs useful?)

At 12:49 AM 5/16/97 +1000, Rick Jelliffe wrote:
>Perhaps because many people (me too, incidentally), having successfully
>not used PIs, realise that it is just an accident of the tools and jobs
they have
>done, and that others in the same industry use them all the time.

I've used PIs frequently to pass information along through transformations
at various points and found them to be very useful only because I didn't
have to manufacture and manage a set of DTDs just to match the
pseudo-structure that I created along the way. It was easy, it was
convenient, and it was a closed loop, they weren't retained as persistent
data.

ArborText uses PIs to good effect by saving edit-state information in the
document including the last cursor position, navigational aids, and
whether or not an element is "collapsed" to name but a few. I think
that these are excellent uses of PIs. In fact, I'd like to see some
of the edit history saved in the document also so that I could still
"undo" the last few edits on a newly-opened document. I may not like
the formatting information that the editor also puts in PIs, but it
doesn't bother me all that much either.

PIs are an escape hatch and their contents is open to the imagination.
Personally, I like escape hatches. They make life easier for me. But
then, I use them in a very disciplined manner, always remaining cognizant
of the fact that I am acting "outside the law" as it were.

Should XML have PIs? Personally, I don't really care, but only because I
don't see any benefit in using XML as a source format for the bulk of the
information that I have to worry about. However, others who choose to use
XML as a source format may want or even need this capability.

Should the internals of PIs be structured? No. Leave them alone. If we
start definining the internals, it will only be a severe distraction and
we will start to give them validity as "first-class" markup and increase
the likelihood of misuse. People should be free to stuff what they want
in a PI, but they should *never* have the expectation that the content
of a PI is interchangeable.

As long as they aren't in the way of something critical, leave PIs in,
but leave them alone. I'd rather have them disappear than have them
"interchangeable."

bob

Received on Thursday, 15 May 1997 19:58:44 UTC