Using CVSup

Tue Jun 27 16:28:04 UTC 2000

On 27 Jun 19100, Forrest J. Cavalier III wrote:

> [Russ, did Barry's message go to inn-workers?  I never got it.]

Pr'bly not, I didn't see it myself, but I've had something like
1/3 to 1/2 success rate with my recent postings

> The URL may be misleading but the CVSup checkout tag used is '.',
> which means the latest and greatest.  (At least the latest!)

Thanks!  But actually, the truth is that I wanted to look at one
of the source files in -STABLE, to see how it handles the case I
ran into a few days ago under inn-STABLE_2_2-1999-05-31_03-01
where the active file happened to be exactly a (mumble)-times
multiple of the OS page size, and actived died, and neither it
nor nnrpd's fallback were able to cope with it.

> (Knowing Barry's tendency for local hacks, he will not want to use
> HTTPsync, since it replaces local changes.)

Bah, local hacks are history.  Or at least too much work for my
replacement to bother with, and don't lend themselves to tagging
along behind a constantly-evolving branch.  I haven't even tried
to merge in my 1.5 hacks that are currently running on our world-
facing peering boxes into the 2.* machines that are hidden from
the world, but which could benefit from the additional innreport

What I'll try to do is get as many hacks as possible merged into
the official source, wrapped with inn.conf wrappers when needed,
but once it comes time to convert my playground into a production
machine, it'll have to be pure unadulterated virginal source material,
otherwise a maintenance nightmare will rear its ugly head.

With that, there are a few more hacks I've added that I should
properly wrap with inn.conf selectables, like this:

(This makes use of the Virtual Path I've set and may well be too
tied to it in my own k0deZ)

Message-ID: <cAsCaDe.CAScADE.casCade.MEOW.MEOW.meow-100000 at>

I'm happy to see that some (even as trivial as that above)
`encryption' of the Hosting-Post header is being discussed...

Newsgroups: dk.edb.internet.udbydere
This.  To indicate that the article was posted on this machine,
I've appended .POSTED! to the VirtualPath; anything following
is user-supplied.  This is a common feature of The Enemies'
software packages.  It probably should be an inn.conf toggle.

Reply-To: pedophile at
X-Trace: 962107431 eeb969c6acf948fa6d673e41f94faa80

Again, `encryption' of the X-Trace header has been mentioned.
I just threw this together to `hide' customers from 'bot scans
made on newly-posted articles.

Another anti-NewsAgent hack that I had been running could have
been helpful a day or two ago.  I do my best to filter out rogue
NewsAgent cancels and supersedes and other messages, but some
do slip through.  To try and minimize the damage done by floods
of supersedes, I turned off the ARTcancel() part of things long
ago and logged the message (which I went ahead and wrote, so a
perl filter that either passes it or drops it doesn't help)...

Jun 27 12:26:23 innd: SERVER ignoring supersedes
Message-ID <n0avgyvh12p.fsf at>
Jun 27 12:27:19 innd: SERVER ignoring supersedes
Message-ID <Qzimbabwe-vote-resultURgIm_AuR.RMi9_AuR at>

Now the bulk of these are ClariNet posts, which we shouldn't be
seeing but they leak like a rusted-out sieve, and enabling the
inn.conf option I added will both leave the old article on the
spool (for safety) as well as write the new one.
ignoresupersedeshdr:    true

Is this something that people would want to have available to use?
And fer cryin' out loud, why can't I come up with a better name?

Another thing that I wanted is an inn.conf and/or readers.conf
way to override what looks to be a hardcoded 10 second timeout
for nnrpd reader connections, which is hardly enough time for me
to type the full name of my favourite pr0n froup.  I haven't
written this yet.  Would this be a useful thing to have?

If any of these things are in -CURRENT already, my apologies,
since I'm only playing on -BETA for the future production box.

More information about the inn-workers mailing list