Tim Bray wrote:
On May 15, 2005, at 9:43 PM, Robert Sayre wrote:
<evilExtension /> Legal? Which part of the spec rules it out?
No part. Per http://www.atompub.org/2005/04/18/draft-ietf-atompub-format -08.html#rfc.section.6.2 it's unknown foreign markup, with clear rules for how to handle it, right?
Yes, there are clear parsing rules. What's the benefit of allowing such markup?
The benefit the Web derived from HTML's implicit-but-universally-implemented MustIgnore rule; it introduced enough slack into the system that people could experiment without breaking things.
Related resource: http://www.webratio.com/images/20050408Bosworth.pps
More generally: ruling things out should be avoided unless (a) you're really sure they're harmful and (b) you think you can actually successfully enforce it. -Tim
A concrete example of the implications of such a rule:
http://diveintomark.org/archives/2003/04/13/object_and_internet_explorer
... in particular, follow this link:
http://www.robinlionheart.com/stds/html4/objects.html
The only way that authors of future revisions of the specification can be confident that they are adding things that don't break anything is if either (1) some names (e.g., the atom namespace) are reserved for such things, or (2) all future additions go into new namespaces.
I do believe that schemas and/or the feedvalidator can enforce such a rule. And I don't believe that a requirement that people who wish to extend Atom do so in their own namespace is overly burdensome.
- Sam Ruby
