There has been discussions or allusions to POSTing invalid Atom Entries, particularly entries without atom:id, atom:author or atom:link (atom:[EMAIL PROTECTED]"alternate"] ?).
Let's consider each element in some more detail.

1. atom:id

"tag" URIs are relatively easy to generate [1], provided that you have the "authority" part. This could eventually be guessed, but not always (Robert doesn't own www.blogger.com [2]). However, it's a really simple configuration parameter.

The window below could be shown when the user first creates an entry and hasn't yet configured the "tag-uri authority".



Extensions to the protocol might add an atom:id generator, an atom:id template, a sentinel atom:id given in the introspection document, or maybe a combination of atom:id template and sentinel atom:id (the template, when processed, gives you an UUID, and the server can eventually replace it with an UUID of its own; this works only if it is clear that this might be a violation of AtomFormat).

2. atom:author

Who knows who is the author of an entry more than its author, so the AtomPP client?

With a non-authenticated collection (e.g. comments collection), someone willing to stay anonymous could send an empty atom:author/atom:name, or <atom:author><atom:name>Anonymous</atom:name></atom:author>.

Implementations might “extend” the protocol by replacing <atom:author><atom:name>[login of authenticated user]</atom:name></atom:author> with predefined information concerning that particular user (and eventually reject the entry if the atom:name is different than the username of the authenticated user).

3. atom:link

I really don't see the problem here…


[1] http://www.imc.org/atom-protocol/mail-archive/msg02691.html
[2] http://www.imc.org/atom-protocol/mail-archive/msg02733.html

--
Thomas Broyer

Reply via email to