OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [xml-dev] RE: Namespaces Best Practice

David Brownell  wrote:

> > >
> > > In every style guide I've ever seen, the guidance for lexical scoping
> > > mechanisms is to use them to localize information to its most natural
> > > scope.  Loop-private variables are invisible outside the loop, etc.
> >
> > For the record, this best practice is lifted from "Common XML" itself a
> > practices guideline:
> > http://www.simonstl.com/articles/cxmlspec.txt
> That writeup adopts a per-file definition of "most natural scope",
> which is more extreme than any definition I'd adopt, though
> it fits into the "wriggle room" which I mentioned.
> That's a bit less extreme than what you wrote, though
> that still wouldn't make me adopt it! :)

I really wasn't intending to be extreme, nor to indicate that this "best
practice" is intended in any way to be some sort of law that must be
followed, rather it is intended simply as a "best practice" to be used as a
guideline. I don't myself follow these guidelines in all cases, but do try
to adhere unless there is a good reason not to. These practices are in no
way intended to supercede the XML Namespace recommendation itself, rather to
provide guidance on usage of this rec in ways that tend to avoid problems.

> > I fail to see how a "loop private variable" relates to a namespace
> > by a URI. This is like comparing apples not with oranges but with entire
> > countries which themselves have farms which themselves contain orange
> > groves.
> That's like ... hyperbole!

The relationship is: "loop private variable" denotes a transient value that
most optimizing compilers optimize out of existence. In contrast a namespace
name is a URI. A URI defines a point in web space, one that itself may be
associated with all sorts of individual resources. So the analogy is between
a single thing and a potentially complex container of things.

In terms of such declarations, an XML document provides for a DTD at the top
of the document, not fragments of DTD intermixed with XML. The so-called

> The relationship is: both bind values to names.
> Both use "lexical scoping" to determine values at use point.

I see where you are coming from. During parse phase a variable is not
typically bound to a value. Of course locally scoped variables may often be
replaced with constants during an optimization phase but only in certain
circumstances. Frequently the values of a program variable are only
determined when the program is actually run.

On the other hand, I am strongly suggesting that the URI bound to a prefix
ought not change. In any case this value is always available at parse time.

Perhaps this may make this more clear: XML Documents should be designed to
allow a human to easily determine the namespace URI associated with an
element prefix.

It seems to me that placing the prefix-URI bindings at the top would ease
this task, but if you would rather perform this somewhere else in a
document, then I see no problem. What ever is more clear for a person to
read. (this -ought- not make any difference to a well designed XML parser so
this guideline exists entirely to allow people, so inclined, to better read
XML documents)

In contrast the value of a variable typically changes as a program is run,
so there is no way to ensure that a human will be able to know the value of
any variable simply by looking at a program -- of course you may have more
intrinsic Mhz than me, but my own brain it still stuck at less than HP
programmable calculator speeds.

Furthermore, programs may have large numbers of variables and so it is often
of true benefit to declare them close to use. Most XML documents ought not
have so many namespaces and indeed if we start getting into single documents
with hundreds of namespaces, we ought either consider using fewer
namespaces, or changing the best practice at that time.

> Information scoping (one variant is "information hiding") is a classic
> best-practice in systems development.  Lexical scoping isn't only to
> support rebinding of names to values; it's also to minimize the amount
> of global state (vs maximizing it, per the CXML policy or the one
> which you proposed).

Sure. However my intention is to point out that best practice for systems
design is not necessarily best practice for document design. Similarly
minimizing global state is not always the most efficient nor desirable way
to do things. For example in XML parsing assuming that prefix-URI bindings
always occur at the root, this will minimize the number of namespace binding
and unbinding events.