[ietf-nntp] OVER message-id

[Ken Murchison]

Andrew - Supernews wrote:
>>>>>>"Russ" == Russ Allbery <rra at stanford.edu> writes:
> 
> 
>  Russ> Andrew's cluster has an additional problem that Xref data isn't
>  Russ> actually stored in the article,
> 
> Actually I do store it in the article in the system we currently have
> live, though that will no longer be true in our next major release.
> 
> I was in fact describing the Diablo cluster design (which is
> increasingly widely used by ISPs and newer entrants to the commercial
> Usenet market) when I said that Xrefs were not stored in the article.

If a server doesn't or can't easily store the Xref header in the 
article, can it provide it in the OVER response, but not in a 
HEAD/ARTICLE response?

I'm asking the question for my own purposes.  Because my Cyrus NNTP 
server is layered on top of an IMAP infrastructure, I can't easily (or 
not at all) store an Xref header in the articles.  I *can* generate the 
Xref content after the fact and provided it where needed.  If I insert 
an Xref header in the HEAD/ARTICLE response, then I also need to adjust 
the :bytes and :lines metadata on the fly to account for the Xref 
content.  If I just provide the Xref content in an OVER response, I 
don't have to munge anything else.

Along these lines, does Xref actually *need* to be a header, or can it 
be treated as metadata?  Obviously some/most servers will store the Xref 
content in the article as a header, but do clients really care where the 
info is, as long as they can get it in an OVER response?

-- 
Kenneth Murchison     Oceana Matrix Ltd.
Software Engineer     21 Princeton Place
716-662-8973 x26      Orchard Park, NY 14127
--PGP Public Key--    http://www.oceana.com/~ken/ksm.pgp