clewis@lsuc.uucp (Chris Lewis) (12/10/87)
Finally! C-news at LSUC seems to be solidly up and running.
It has run for about a week without any problems. The previous
week and a half were ... interesting. The installation is
complete except for some minor modifications we wish to make to
it to make administration (and bullet-proofness) a little easier.
All downstream sites are using the C-News batching system. However,
we might in future convert over to my "batcher" control system I posted
some time back - it's a lot faster. After I upgrade it's batch formatting
to be as flexible (or, at least a little easier to "flex") as C-news's
batcher.
If anybody is encountering any problems with articles passing thru here,
please contact me (clewis@spectrix) and/or Dave Sherman (dave@lsuc).
I won't be on this system as much as I was over the last month.
A couple of random notes on the installation for those people interested
in bringing up C-news:
1) Unless you want to spend a fair bit of time hacking this thing into
shape, I would (unfortunately) recommend that you don't install it.
Think of it as a "Hacker's news system" (ala "Mechanic's car").
There's a lot of little gotchas (eg: permissions, and exact locations
of some of the things, and how things go together) that will probably
cause some interesting evenings.... If nothing else, it ain't as
easy to install as later versions of B-news.
2) Some portions of C-news are REAL fast. Once that relaynews program
gets around to eating a batch you won't believe how fast it processes
articles. HOWEVER, that isn't the whole story.... On lsuc, it takes
at least two *minutes* for an article to hit the spool area after
it's been posted locally (elapsed time from "send" to its appearance
in the log file). So, don't sit there waiting for your cancel command
to kick an article out from under you.... Also, the multiple layers
of shell script wrapped around relaynews causes long delays between
relaynews being able to eat articles (though, of course, a lot of
that is due to uncompression...).
3) Make sure you've fixed the returns in putenv.c (if you're using it).
REAL mysterious relaynews failures.
4) You might want to make a change to realrnews (or is it servernews) to
touch incoming/stop whenever relaynews returns a non-zero return code.
This will stop incoming stuff dead til you remove it. Upon certain
occasions, not doing this has caused the failure recovery code
to fill up /tmp and my mailbox with the contents of every batch we
received. Blew lsuc out of the water twice.
5) The documentation is, um, sparse.
6) The batcher is slow. (though, so's this machine.)
Think of C-news as a prototype. Works, but needs some tinkering. Didn't
have to touch any C code except for the putenv problem. Though, there
are a number of things that could be put into the programs to make them
a little more bullet-proof (eg: ensuring that they're being run as
the right userid).
C-News has a number of very helpful administrator-nice features that make
it operationally nicer than 2.11 (except for some of the failure recovery).
The next version of C-News should be really something.
If anybody's interested, I'll post my more complete list of problems and
suggestions....
--
Chris Lewis, Non-resident C-news Hacker,
Real: {uunet!mnetor,utcsri!utzoo,ihnp4!utzoo,utcsri!utzoo}!spectrix!clewis
Virtual: {same as above}!lsuc!clewis