From: Charles Lindsey (chl@clw.cs.man.ac.uk)
Date: Tue Sep 05 2000 - 05:02:42 CDT
In <ylhf7v4u9m.fsf@windlord.stanford.edu> Russ Allbery <rra@stanford.edu> writes:
>Charles Lindsey <chl@clw.cs.man.ac.uk> writes:
>> Russ Allbery <rra@stanford.edu> writes:
>>> and just use the colon as part of the value since it doesn't change.
>>> If whitespace is allowed around the equal sign, that can be dealt with
>>> although it's annoying. Comments are right out. If we're going to
>>> allow spaces in posting-host, then it should *always* be quoted so that
>>> the parse can be easy; sometimes quoted and sometimes not quoted is
>>> much more annoying to deal with.
>> The whitespace and comments are there because they are there in all our
>> headers (except Newsgroups and Path, which we singled out).
>Right, I know, but it makes the header much more difficult to parse. If
>you want to replace something that's working quite well for the majority
>of people, you need to make it *really* easy to use the new thing.
>> Indeed, comments regularly appear in X-Trace.
>Practically no one filters on X-Trace; the only use X-Trace really has
>right now is to provide information to the originating site. Everyone
>filters on NNTP-Posting-Host because it's considerably easier and most
>sites that generate X-Trace also generate NNTP-Posting-Host. Besides, the
>comments in X-Trace for *most* servers are reliably always in the same
>place (and the fact that they aren't that way everywhere is a big part of
>the reason why no one filters on X-Trace).
>> The alternative is to pass the whole header through an
>> "unfolding/de-commenting" process before trying to parse it. This is
>> probably a useful tool to haver available for other headers, too.
>Yes, but you're putting more programming onto the filter writers and
>they're under no obligation to pay any attention to our weird new headers
>at all. Unless it's easy to use, I don't think anyone's going to use it.
Filtering is not a perfect art, so a few misses are to be expected.
In practice, ISPs are not going to be inserting comments in stupid places,
so filtering on
Injector-Info:.*posting-host="?..."?
will work often enough to be useful.
>> I was trying to have a very specific syntax for each parameter (see my
>> reply to Clive). So "identity" would be a different header with a
>> different syntax. Remember that all these headers are optional, so the
>> injector includes just those which it considers appropriate to its
>> situation.
>That's going to make it basically worthless for spam filtering, then, and
>everyone's going to just keep using NNTP-Posting-Host. I know I would.
I don't see why that makes it worthless. If the "posting-host" parameter
is present and is useful, then use it. If that is not useful (because of
too many false positives) then an additional "identity" or "user"
parameter might provide additional filtering possibilities.
The main thing is to make sure that each parameter has a simple and well
understood syntax that is easily parsed. Trying to cram everything into
one parameter just makes parsing more difficult.
-- 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