From: Charles Lindsey (chl@clw.cs.man.ac.uk)
Date: Wed Aug 18 1999 - 05:30:26 CDT
In <199908172151.XAA19353@kairos.algonet.se> sommar@algonet.se (Erland Sommarskog) writes:
>Charles Lindsey <chl@clw.cs.man.ac.uk> writes:
>> 5.6.2. Adding a path-identity to the Path header
>I'm still not happy with this section.
>> The path-identity added MUST be unique to that agent. To this end
>> it SHOULD be one of:
>Only SHOULD?
I agree with Brad. The prime purpose of Paths is prevention of loops, etc.
Uniqueness is a MUST, how it is achieved is a matter of convenience.
>> 1. A fully qualified domain name (FQDN) associated (by the Internet
>> DNS service [RFC 1034]) with an A record, which SHOULD identify
>> the actual machine prepending this path-identity. Ideally, this
>> FQDN should also be "mailable" in the sense that it enables the
>> construction of a valid E-mail address of the form
>> "usenet@<FQDN>" or "news@<FQDN>" [RFC 2142] whereby the
>> administrators of that agent may be reached.
>>
>> 2. A fully qualified domain name (FQDN) associated (by the Internet
>> DNS service) with an MX record which MUST enable the
>> construction of a valid E-mail address of the form
>> "usenet@<FQDN>" or "news@<FQDN>" whereby the administrators of
>> that agent may be reached.
>I have to admit that the concepts of A and MX records are foggy to
>me, but I fail to see why there are is such difference in strength
>between the call for usenet@<FQDN>. In point 1. it's the very mild
>"ideally, this FQDN should..." (lowercase "should"!), and in point 2.
>it's MUST. Is there something I miss? Could we not condense this into
>one point:
My present text is essentially where our discussions had got to. It seemed
good to separate the A and MX cases so we could get a clearer picture
ourselves. Now we can see that they are somewhat similar, we can review it
again.
There is, however, still a difference. Suppose isp.com relays through a
machine news17.isp.com (it has at least 17 relaying machines,
apparently). There is an A record for news17.isp.com, but it is too much
hassle to make all 17 machines mailable (mail should be sent to isp.com,
and most newsadmins who want to contact them could probably guess that).
Alternatively, the article might pass through a chain of machines at
isp.com, all with A records, but we do not insist that all are mailable
(though one would hope that at least 1 of them was).
OTOH, isp.com MUST NOT, in that case, inject at news17.isp.com (though he
could add two path entries "news17.isp.com/isp.com%" to make it legal).
> 1. A fully qualified domain name (FQDN) associated (by the Internet
> DNS service [RFC 1034]) with an A or an MX record, which SHOULD
> identify the actual machine prepending this path-identity. The
> addresses "usenet@<FQDN>" or "news@<FQDN>" [RFC 2142] to reach
> the administrators of that agent MUST be valid.
No, an MX record does not identify an individual machine, and mailability
is perhaps less important than identifiability, except at injectors. So I
think my two paragraphs now express the correct semantics, though they
might indeed be simplified once we are agreed on that.
>> Of the above options, nos. 1 to 3 are much to be preferred, unless
>> there are strong technical reasons dictating otherwise. In
>> particular, the injecting agent's path-identity MUST, as a special
>> case, be an FQDN mailable in the sense defined under option 1, or
>> with an associated MX record as in option 2, and it MUST be
>> followed by the special delimiter '%' which serves to separate the
>> pre-injection and post-injection regions of the Path-content. See
>> the Duties of an Injection Agent (section 7.1). In the case of a
>> relaying or serving agent, the delimiter is chosen as follows.
>This is also funny, because it first says that UUCP names are among
>the preferred ones, but in the next sentence declares them third-
>class citizens. Given the collapse of options 1 or 2 above, this
>could be:
Well they are indeed third class insofar as you MUST NOT inject at them
(because most people don't know how to mail to a UUCP address anymore :-(
). But you might indeed argue that forbidding UUCP-site-injection is going
to far, but that is what we seem to have agreed to.
[snip]
>Here I have also removed the extraneous "unless there are strong....",
>and split into two sentences and paragraphes two quite different
>issues with injecting agents.
I think your detailed changes here must await decisions on combining
options 1 and 2, but in the meantime I like your paragraph splittings, and
have adopted them (more or less).
>[I still don't really see why option 4 could not be comibined with
>a usenet@ address, but I rest that case.]
It is not forbidden, but less likely to happen.
However, I would like someone to comment on the fact that I have removed
entirely the possibility on enclosing naked IP addresses within [...].
-- Charles H. Lindsey ---------At Home, doing my own thing------------------------ Email: chl@clw.cs.man.ac.uk Web: http://www.cs.man.ac.uk/~chl Voice/Fax: +44 161 437 4506 Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K. PGP: 2C15F1A9 Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5