Re: Event handling

When I read this, I quickly started to remember systems that
sarted out without a desire to "require (or even encourage)"
name scoping.  Nightmares arose before long.

So I think it's very appropriate that W3C provide basic
guidance about how to avoid name collisions.

Now, I recall providing that feedback to earlier drafts with
such comments.  I was even told that W3C-defined events would
have unique prefixes (except "level 0" compatibility ones).

But that issue still hasn't been resolved -- or, as I recall,
even mentioned as one of the issues raised by the public
feedback cycle.  I'd call that a clear W3C process failure.

Note that this is the same issue that came up with discussions
about reserving other namespaces in DOM, for node types and
exception codes.  No W3C spec should _ever_ just define a new
namespace and not define the policy for managing it.  In this
case, some reasonable policies might be:

  (A) Only W3C can ever define event types (undesirable).

  (B) Certain prefixes are reserved to W3C, and all others
      are reserved to applications.  (Undesirable, since it
      offers no policy for infrastructure libraries built on
      top of DOM...)

  (C) Like (B) except some policy for infrastructure layers
      is defined ... e.g. reversed domain names.  (preferred)

  (D) Event types (names) have an associated namespace URI,
      including the ones in the DOM spec.  (OK, but complex.)

The DOM WG has precedents in its own history in this space:
the way that some implementations defined new node types and
exception codes, thus causing interoperability problems.  It
does NOT need to recreate such problems.

- Dave


----- Original Message -----
From: Patrick Schmitz <pschmitz@microsoft.com>
To: <keshlam@us.ibm.com>; <www-dom@w3.org>
Sent: Thursday, March 16, 2000 4:06 PM
Subject: RE: Event handling


> > -----Original Message-----
> > From: keshlam@us.ibm.com [mailto:keshlam@us.ibm.com]
>
> ...
> >
> > It might be appropriate to encourage folks to manage the
> > event-name space
> > as URIs or via reversed-domain-name prefixes or something of
> > that sort, to
> > reduce the risk of collision.
>
> In many cases, we expect authors to define events that are associated with
> specific media and documents; they may not want the overhead of such a
> naming scheme.  I see your point, and folks are free to take that approach
> where it may make sense, but I am not keen to require (or even encourage)
> namespace qualified event names.
>
> Patrick Schmitz
> Program Manager - Web Multimedia Standards
> Microsoft
>

Received on Monday, 20 March 2000 14:21:12 UTC