[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
draft-allbery-usefor-usepro-00 errata
As a summary from the recent long threads, here are the changes I believe
should be made to draft-allbery-usefor-usepro-00 if we go forward with
that document. The first one below has not previously been discussed on
the list.
* USEFOR was changed (correctly, IMO) to allow only WSP between the
arguments of control commands, not FWS. I caught this for newgroup and
rmgroup but missed it for checkgroups and cancel, both of which need
FWS to WSP changes.
* application/news-groupinfo and application/news-checkgroups need
language limiting the selection of character sets. I like Harald's
proposal and would tend towards something like:
Implementations predating this standard may not understand MIME
headers and expect newsgroup names to be in ASCII. Therefore,
regardless of what charset is used, the result of reading each octet
of the body and setting bit 8 to zero MUST be a valid message
specifying the same newsgroup name [or names for checkgroups]. There
is no requirement that the newsgroup description survive this
treatment.
* multipart/related in newgroup control messages should be
multipart/mixed instead.
* A separate sub-section of duties, before the individual agent duties
and possibly as part of the same sub-section as the identity
discussion, should be added specifying Path header construction. The
Path header example should be moved to be a sub-section of that
section. That section can lay out all the construction requirements
for the Path header field and just be referred to by the individual
duties sections.
* The possibility of adding multiple path-identities should be
reintroduced as part of the rework of the Path description.
* Serving agents also are not required to modify the Path header field if
processing an article from a relaying agent or injecting agent that's
part of the same server. This can be handled more generally in the
redone Path section.
* application/news-transmission should explicitly note that, contrary to
the previous registration, batches are not permitted.
* The trimming requirement for References should probably go back to
MUST.
* The correct description of special routing for newgroup and rmgroup
control messages is something more like:
Exceptionally, control messages creating or removing newsgroups
(newgroup or rmgroup control messages, for example) SHOULD be relayed
if the affected group appears in its Newsgroups header field and the
sending agent would have supplied and the receiving agent would have
received the newsgroup affected by the control message had it
existed, even if it currently does not.
The current USEPRO text for construction of the Newsgroups header field
for those control messages should be restored (and probably as a
general statement about group control messages that affect only one
group rather than stated independently in both newgroup and rmgroup).
* We need some resolution of the different conflicting syntaxes for ihave
and sendme control messages. Path of least resistance is probably to
revert to the current USEPRO tactic of two separate ABNFs.
* (Still under discussion.) We may wish to reintroduce the possibility
of a path-identity that is not a resolvable name in DNS.
* (Still under discussion.) We may wish to add a sentence about the
construction of newsgroups in the to.* hierarchy.
--
Russ Allbery (rra@xxxxxxxxxxxx) <http://www.eyrie.org/~eagle/>