From: John Stanley (stanley@peak.org)
Date: Sat Apr 21 2001 - 22:31:55 CDT
Erland Sommarskog (sommar@algonet.se):
>But then it is clear where the problem arises: when Site B or whoever
>defined the header said "this is a Local Header according to the
>definition in Usefor".
Let me explain what NoCeM is, dear. It is a message sent by certain people
to indicate that another message or group of messages might be ones that
should be removed by a server. The decision to honor a NoCeM is ENTIRELY
LOCAL. What a site does with NoCeM messages is up to them, and them alone.
If a site uses NoCeMs to insert a header in the articles they serve so
that users may filter on NoCeM status, then that decision is also ENTIRELY
LOCAL. Each admin has the right to decide for his site, but has no right
to decide for others, that the information in the NoCEm is significant to
any other site. That's what makes any such header a local header, and it
was CORRECT for the people who wrote the system being used by Site B to
define it as a local header. They defined it this way based on the
political side of USENET management, not Site B as you previously claimed.
(And I'll point out that the "definition" of "local header" is not a
definition but a statement of one property that it has, in subjective
terms. If you don't understand the difference, think about the value of a
dictionary that has the entry "apple: a red thing, and you find them in
trees.")
>Then they also said "This header MUST be removed by other sites".
The people who defined the header as local DID NOT say that it must be
removed when passed to another server. That's simply not true. THIS GROUP
wrote the draft which says this. It does not say MAY remove, it does not
say SHOULD remove, it says MUST, as if there were some serious
interoperability issue created by this header which nobody knows exists
today.
I suggest you read the draft before you make daft statements about who
said a header MUST be removed. It wasn't Site B, Site C or the authors of
the NoCeM code.
In fact, here's the section of the draft, because I just noticed that it
is even more absurd that I initially considered.
] Headers with the local property are significant only to a particular
] serving agent (or perhaps a cooperating group of such agents). They
] MAY be removed by relaying agents before propagation, and MUST be
] removed (and replaced as necessary) by serving agents when received.
So, in our draft, we say that local headers may be of interest to a
"cooperating group of [serving] agents" but then say that they MUST be
removed by "serving agents" when received. So we admit that groups of
servers can cooperate on local headers, but then say that the header they
are cooperating on MUST be removed when a serving agent receives an
article. That's a direct contradiction. If I didn't know better, I'd think
the person who wrote this paragraph was loopy.
>But the actual problem is really much smaller. Because even though Site C
>MUST remove the header, they are permitted to insert its own local header,
>and provided it is not precluded by the documents that define your
>hypothetical nocem-header, they could take the value from the header they
>just removed.
Then they haven't removed it, have they? And if they had the resources to
run the code that Site B does, they would, but they don't, so they cannot
create their own header. They went into this saying they would cooperate
with Site B, but this draft does not allow them to.
>Usefor sets the definition. Later RFCs or similar documents are free
>to use it, but they should of course know what they are doing. Thus,
>Usefor knows damn well what kind of header this is, because Usefor
>gave the definition of a Local Header.
USEFOR knows it is a local header, but like I said, you have no idea what
this header is or what it does. Yet you will make it mandatory that it is
removed. This is a clear violation of "if you don't understand it, leave
it alone" and RFC whatever where the use of MUST in RFC is discussed. To
claim that you "know damn well" what this header does before it has ever
been defined is insulting and ridiculous. Maybe it's your belief that you
are in the "real world" and others here are not that misled you into
making such a statement. I don't know.
The only header that meets the criteria of being local and causing
problems is XRef. If XRef is the problem, then outlaw passing XRef around.
But then, cooperating servers may want to share that if they share a
spool. But another then, is it really necessary to say anything about Xref
at all? I mean, software authors will find out pretty quickly that their
code doesn't work well if they keep XRef headers sent to them by other
sites, and they will probably be creating their own local copy anyway.
If you say that Local headers MAY be removed, you've given them permission
to make their code work and not have prevented other Local headers that
can be useful from being useful. You've also not directly contradicted
yourself in two sequential sentences in the same paragraph. Sheesh, if
saying the same thing in two [separate] places is bad, how bad is it to
say different things in the same place?
And the old same place is out back, here's the key, clean up after
yourself.