Article View: demon.ip.developers
Article #623Re: KA9Q mailkill can kill wrong messages.
From: Robert Mortimer
Date: Sun, 23 Mar 1997 00:00
Date: Sun, 23 Mar 1997 00:00
53 lines
3060 bytes
3060 bytes
In article <858683625snz@wash.demon.co.uk>
john@wash.demon.co.uk (John Washington) writes:
> On Monday, in article <s858637073snx@warren.demon.co.uk>
> leveret@warren.demon.co.uk "Nick Leverton" wrote:
> NL> There's a bit of an "oooh, nasty" in KA9Q's mailkill handling. The
> NL> variable 'bad_sender' in smtpserv.c is global (only one copy for the
> NL> system), where it should be locally allocated in sptpserv() (one per
> NL> receive process). This means that, when you have two receives going in
> (and 2 recvs is quite frequent for me, I don't know if that's true
> for everyone)
Doesn't seem to happen that often for me.
> NL> parallel, a rejection on one will also cause the other, valid, stream to
> NL> perform a mail rejection, if it is itself inbetween the "MAIL FROM" and
> NL> "DATA" stages, inclusive, of the protocol.
> Ouch. And what about the other statics alongside, 'fail', 'default_address'
> and 'Ssmtp'? Looks easy to move 'bad_sender' into smtpserv(), but the
> others are used by more than one function.
I changed the fail one quite easily for DNOS, I'll have to have a look
at the others.
[snip]
> NL> to track through all the implications of this. Note that there is
> NL> another independent variable also called fail in mailit().
> Yuck, yuck, YUCK. Grrrr.
> And, while we're discussing yucky coding style, someone since
> 2.12 has decided to lay out the smtpserv.c code to his
> personal taste, including indenting the function names. By
> doing this he's broken the convention in ka9q source that
> has all function definitions starting in column 1, so that
> you can grep for 'funcname' at the start of the line and
> find that definition easily.
Horrible isn't it it makes following the code a lot harder.
> Talking of mailkill, what would be really nice would be some
> scheme whereby additions could be made automatically.
> I'm imagining some group like demon.mailkill where people
> would post addresses of recent junk, and ka9q would
> autoextract and update mailkill. Or some agreed format, like:
Trouble is it would be very easy to abuse unless it was moderated
and how can the moderator tell if that person has really sent spam/uce
or not?
[snip]
--
Robert Mortimer
"Trust the computer The computer is your fiend"
http://www.maghmell.demon.co.uk/dispc/index.html
Technical Details
Message-ID:
<859186485snx@ogham.org>Path:
arc1-ca.pugleaf.net!unknown!not-for-mailReferences:
<s858637073snx@warren.demon.co.uk> <858683625snz@wash.demon.co.uk>Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 10bb05,4a438dcc7e2e89ee X-Google-Attributes: gid10bb05,public X-Google-Thread: fb525,4a438dcc7e2e89ee X-Google-Attributes: gidfb525,public From: Robert Mortimer <robm@ogham.org> Subject: Re: KA9Q mailkill can kill wrong messages. Date: 1997/03/23 Message-ID: <859186485snx@ogham.org> X-Deja-AN: 227829239 References: <s858637073snx@warren.demon.co.uk> <858683625snz@wash.demon.co.uk> X-Mail2News-User: Giolla@ogham.org X-Mail2News-Path: office.demon.net!maghmell.demon.co.uk Organization: Giolla Decair Designs Reply-To: robm@ogham.org Newsgroups: demon.ip.developers,demon.ip.support.pc