Re: #1047 Path field delimiters and syntax - status

["Charles Lindsey" <chl@xxxxxxxxxxxxxxxx>]

In <D52515DCC877BF7A1886A6AB@xxxxxxxxxxxxxxxxxxxxxxxxxxxx> Harald Tveit Alvestrand <harald@xxxxxxxxxxxxx> writes:

>At the moment, I have

>- My suggestion to separate "diagnostic information" from "identities"; in 
>the syntax, the result is that parsers need to expect both.

This is where USEPRO and USEFOR interact, so we need to fix both together.

At present, the word "diagnostic" is not used, so it would need
definition. The only examples of "diagnostic" mentioned to date have been
where an agent inserts into the Path the observed origin of an article
(which may be different from the claimed identity inserted by the sending
site). Are there any other "diagnostics" people have in mind? Currently,
the only example permitted is in conjunction with the MISMATCH keyword,
but I think we are agreed that other cases should be explicitly allowed
(with or without extra syntax such as SEEN/MATCH/MISMATCH).

>- A lot of debate on where and why IP addresses occur in the path, 
>including some ideas from Charles about annotating the IP addresses with 
>extra syntax that allows some more intelligent guessing on why they were 
>added

>My current suggestion for SYNTAX (that is, what goes into USEFOR)

>- Specify "identities" as domain-names or barewords, period

>- Specify that "diagnostics" are PERMITTED in the path, and that this 
>standard does not restrict the format of diagnostics very much (in fact, 
>permit all of [A-Za-z0-9.:], and then shut up)

But that allows ":" in all sorts of places where they are not needed. The
beauty of your earlier proposal was that ":" could ONLY occur within an
<IPv6Address> (and I think we are agreed that IP-addresses are going to be
present, at least as diagnostics, in which case there are going to be
<Ipv6Addresses> whether we like it or not).

But if <Ipv6Address>es are going to be present, then the <bareword>s
"dead", "beef", etc. are going to cause trouble with agents that still
think ":" is a delimiter (not that I think this "trouble" is a serious
showstopper). But if you allow ":"s anywhere within diagnostics, then the
"trouble" (such as it is) could also occur with innocent <bareword>s such
as "wolf" and "demon".

>- Permit : in "diagnostics", not in "identities", and add some appropriate 
>note with warnings on what will happen to DEAD:BEEF:CAFE

But given that the "dead/beef" problem has not been eliminated by your
proposal, what do you actually gain by forbidding IP-addresses in
"identities"? The cat is already out of the bag. You are bolting the
stable door after the horse has bolted (so I shall nip it in the bud :-)
). [Sorry, joke might be hard for Norwegians, but Englishmen with a
smattering of Irish politics should get it.]

I grant you that, if IP-addresses are permitted in identities, there are
grounds for deprecating them (as there are for <bareword>s), but not at a
level justifying "MUST" wording.

And note also that an <Ipv4Address> is a <bareword> :-( .

>Does this make sense?

Not really.

-- 
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Tel: +44 161 436 6131 Fax: +44 161 436 6133   Web: http://www.cs.man.ac.uk/~chl
Email: chl@xxxxxxxxxxxxxxxx      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