nntp@tmc.edu (Stan Barber) (09/10/90)
Well, despite a month of testing, there are still some problems with the 1.5.10 release of NNTP. The following are the only two significant problems reported. 1. Defining SETPROCTITLE when running CNEWS is still problematical. The workaround is to NOT define SETPROCTITLE. This will be fixed in 1.6. I had a number of CNEWS test sites. One of those sites reported the problem about a week AFTER the release. Sigh. 2. GNEWS and the new select() based timeout mechanism in 1.5.10 are not compatible. The workaround is to not run 1.5.10. This will also be fixed in 1.6. This one got by because all my beta testers don't use GNEWS. I will be sure that 1.6 has a beta site that does use GNEWS. Watch this space for more developments.
spike@world.std.com (Joe Ilacqua) (09/13/90)
In article <4094@lib.tmc.edu> nntp@tmc.edu (Stan Barber) writes: >1. Defining SETPROCTITLE when running CNEWS is still problematical. <The workaround is to NOT define SETPROCTITLE. This will be fixed in 1.6. >I had a number of CNEWS test sites. One of those sites reported the problem <about a week AFTER the release. Sigh. How does the problem manifest itself? I am of course going to turn SETPROCTITLE off, but I (and perhaps others) would like to know what to look for... One thing I have seen is nntpd leaving behind unfinished batch files when the connection to the remote is lost. The code seems to suggest nntpd is suppose to clean these up... ->Spike (Spike@World.STD.com) -- "The World" - Public Access Unix - +1 617-739-9753 24hrs {3,12,24}00bps
sob@lib.tmc.edu (Stan Barber) (09/13/90)
The problem with SETPROCTITLE is that it does the environmental save too early. This means that the environment is saved before the code used when POSTER is defined. Therefore anything spawned will run as the wrong user. If you don't define SETPROCTITLE, this doesn't happen. And things work as I had intended. SETPROCTITLE is NOT necessary for NNTP to work properly. It was just a neat way to monitor using ps where NNTP connections were coming from.
nntp@tmc.edu (Stan Barber) (09/13/90)
In article <4094@lib.tmc.edu> nntp@tmc.edu (Stan Barber) writes: >2. GNEWS and the new select() based timeout mechanism in 1.5.10 are not >compatible. The workaround is to not run 1.5.10. This will also be fixed in >1.6. This one got by because all my beta testers don't use GNEWS. I will >be sure that 1.6 has a beta site that does use GNEWS. GNEWS should be replaced by GNUS in the above text. Also, the fix is to detailed below. It is a change to GNUS: A workaround is to have GNUS users put the following in their .emacs files: (setq nntp-maximum-request 1) Apparently GNUS depended on a feature of the server that is outside the protocol spec. Some will argue that NNTPD is broken if it doesn't conform to the spec and others will say it is now broken because it doesn't act like it used to. I am still thinking about this. Please send your opinions to the nntp@tmc.edu mailbox.