* Brian Smith <[EMAIL PROTECTED]> [2007-09-14 19:05]:
> It seems wrong to me to standardize on a mechanism that isn't
> using URIs for identifying entries

The Atom ID *is* a URI.

> especially since the idea of permalinks already exists for HTML
> representations of entries.

Atom Entries need not have a permalink. But they must always have
an ID. Permalinks shouldn’t change but sometimes do; Atom IDs
MUST be unchanging.

We started with the exact same mechanism you suggest:
@rel='in-reply-to' links.

It didn’t work.

We originally said the @href is the Atom ID, but such links are
useless to Atom consumers that don’t know `in-reply-to`, just
like thr:in-reply-to elements are useless to Atom consumers that
don’t know what they mean. So there is no gain from reusing
atom:link with overloaded @href.

Also, sometimes you have both a permalink and an Atom ID. It
would be good to provide both. With a single attribute this is
not possible, so you need two.

The logical thing to do is put the permalink in @href and the
Atom ID in an extension attribute. Then such a link is somewhat
useful even to clients that don’t specifically understand
`in-reply-to`.

Except that doesn’t work. There are two problems:

1. The Atom ID would go in an extension element. But what happens
   when you have no permalink? After all, atom:link REQUIRES an
   @href attribute, and with good reason.

2. You have introduced a bad co-constraint. The link extension
   attribute is only valid on links where @rel has a particular
   value.

You really can’t use atom:link for this.

We tried.

Regards,
-- 
Aristotle Pagaltzis // <http://plasmasturm.org/>

Reply via email to