ietf-nntp HDR

[Clive D.W. Feather]

Stan O. Barber said:
>> * Is it supposed to have one argument or two ? The syntax says one, the
>> text implies two.
> There are two arguments: The header name and the article identifier. I will see
> what needs to be done to clarify.

You need to include both arguments in the synopsis line at the start of
9.5.3.1.

>> * The text "See RFC-1036 for a list of valid header lines" remains
>> confusing. Are we saying:
>>   - the header must be one listed explicitly in RFC 1036, or
>>   - 1036 gives examples of headers you might want to use
>> ?
> 
> Good question. The reference implementation does not care, so it would be
> consistent with the latter  clause.

I think you should go for that unless anyone speaks up fairly quickly.

>> * Must the command work for any header, or can the server reject some
>> headers on the grounds that it doesn't store them in its internal database ?
>> If the latter:
>> - we should say so and note that 503 will be used for unrecognised
>>   headers;
>> - are there any headers that must always be supported (e.g. Subject) ?
> 
> This is confusing to me. Which "internal database" are you talking about? Do you
> mean Overview? Historically, HDR didn't depend on OVERVIEW (and in fact
> preceeded OVERVIEW by many years).

It could mean Overview, or it could mean anything. Remember, this is a
black-box specification. Let me reword the question:

* Can the command restrict the header parameter (other than syntactically)
or is any valid string permitted ? If I write:

    HDR X-Haddock 1-

am I guaranteed a 221 response, or can I be sent a 503 on the grounds that
this server does not support the X-Haddock header for the HDR command.

>> * I don't get the fifth example. Do I have to select a group before using
>> HDR with an article-ID ? If so, which group ? Any group, or must it be one
>> that's mentioned in its Newsgroups line ? This isn't described anywhere in
>> the main text. If the right group has to be selected, then I suggest we
>> need an additional response (and example):
>> 
>>     413  Article not in this newsgroup
> 
> This is a good question. The reference implementation does not require a
> newsgroup be 
> selected first. This example could be eliminated if folks don't want to see it
> work in the manner suggested by the example.

By analogy with ARTICLE/HEAD/BODY/STAT I would say that it should *not* be
necessary to select a newsgroup beforehand. I would suggest wording
something like this (note, by the way, that the word "optional" in the
existing text is wrong):

    ... a period is sent on a line by itself.

    If the second argument is a message-ID, then:
    - If the article exists and includes that header, the server MUST
      return a 221 code followed by a list containing exactly one line.
      If the article is in the current group, the article number on that
      line MUST be either zero or its actual number in that group; if the
      article is not in the current group or if no group is selected, the
      article number MUST be zero.
    - If the article exists but does not include that header, the server
      MUST return a 221 code with an empty list.
    - If the article does not exist, the server MUST return a 430 code.

-- 
Clive D.W. Feather  | Work:  <clive at demon.net>   | Tel:  +44 20 8371 1138
Internet Expert     | Home:  <clive at davros.org>  | Fax:  +44 20 8371 1037
Demon Internet      | WWW: http://www.davros.org | DFax: +44 20 8371 4037
Thus plc            |                            | Mobile: +44 7973 377646