On 7/5/06, Eric Scheid <[EMAIL PROTECTED]> wrote:

On 6/7/06 7:08 AM, "James M Snell" <[EMAIL PROTECTED]> wrote:

> Whether or not to change atom:updated on an update request is
> implementation specific and can only reasonably be determined by the
> server.  A server could choose to use the atom:updated value provided by
> the client, but is not required to do so.

-1. "can only reasonably be determined by the server"??? Servers, generally
speaking, are not imbued with artificial intelligence. Determining whether
something is a "significant" update requires understanding who the audience
is, whether various laws have relevant jurisdiction, and understanding the
motives and agenda of the content producer.

One notion of intelligence would be to ignore attempts to change
updated w/out any material change in the contents of the entry.   The
definition of "material" seems possible  in various domains:  for a
blog, it _might_ means if the PUT doesn't change the content or
summary of the entry.   Adding author info, rights changes,
categories, or link relations wouldn't be considered "material".

I provide the above only by way of an example.   I don't want to get
into the discussion of what *is* a "material" change, I doubt there is
a single definition that fits all potential APP domains (or one that
this group could agree upon :).

Honestly, the case I'm *most* worried about clients that keep bumping
their updated value just to keep their entry at the top of (or
reoccurring in) syndicated views of the collection ordered by
updated... i.e. a form of syndication spamming.   Any spec-language
that keeps the server from preventing such types of abuse is a strong
-1 in my book, making the MAY in James' PACE necessary.

It's naive to think APP clients will always play nicely with others.

-- Kyle

Reply via email to