On Thu, 11 Aug 2005, A. Pagaltzis wrote: > > * James M Snell <[EMAIL PROTECTED]> [2005-08-09 07:25]: > > The second feed illustrates the two forms of the in-reply-to > > element. The dereferenceable form uses the href attribute to > > locate the entity being responded to. > > I am still strongly -1 on making the ID optional. Why do you want > to do that? > > What scenario exists wherein it would be *more* desirable to > provide *only* a dereferencable location but *not* an ID? When > would it be *wiser* to *rely* on a pointer to a resource which is > always in danger of voiding, irrecoverably breaking the > connection from a reply to its parent?
Any scenario where the document being referenced is not Atom. If the comments specification were to restrict itself to just discussion about the Atom community then that's both very insular and hardly likely to encourage anyone to contribute to it. Consider you have a news website that gives you a page with text in it. You want to reply to it. They don't have an Atom feed. Do you give up on ever commenting on their text because you can never reference it ? Do you comment on it but omit any linking details ? Do you complain to the website author to add an Atom feed because otherwise you can't refer to it on your weblog ? There are many document forms which have identifiers to locate them uniquely in the world (Atom id, RSS guid, RFC2822 mid, MIME cid, book isbn, etc), but there are also many document forms which have no unique identifier (HTML documents, PDF documents, physical memos, conversations in the pub (!), newspaper articles, etc). To preclude refering to non-id'd documents would be restricting the specification to only allowing Atom people to talk about Atom people. By all means require any reference to a Atom document to supply an identifier, but requiring it on all references would make it impossible to fulfill on some documents. -- Gerph <http://gerph.org/> ... Want to cry for you. Would it do any good ?
