Ricardo Nabinger Sanchez <rnsanc...@gmail.com> posted ghstqg$qp...@ger.gmane.org, excerpted below, on Fri, 12 Dec 2008 05:45:52 +0000:
> I have been accessing some groups through Shaw NNTP server (Canada), and > it is quite common for messages to be incorrectly marked as read. I > don't see anything special in the logs, and would be glad helping > debugging this as it is very annoying. > > This does not happen on Gmane, for instance, so it might be something > server-side. It is still not clear if this only happens with new > threads or with ones I've already read part of. > > I'm Pan 0.133 on Slackware-12.1 (the package is from Slackware-12.2). > It also "happens" on 0.132, but again, it might be something odd with > the server. The groups I'm following most on Shaw servers are > comp.lang.c, comp.lang.c++, and comp.compilers. We've just been discussing this behavior in a different thread. The problem, I believe, is the intersection of two different things, one that's the news server's problem, one that you can do something about. This is a relatively new hypothesis based on discussion in the other thread and it needs more testing, but it matches the behavior so far observed. The short version is don't use "mark all as read", either in preferences for automatic marking as read (under behavior, either when leaving group or before getting new headers), or manually, using the "Mark selected groups read" function (context menu in group pane or groups menu). There are workarounds... What happens is that some servers deliver messages out of sequence -- the Xref header has a server-specific per-group sequence number which should be always increasing, but sometimes, a server will skip around a bit. This can be due to the size (some arrangements deliver small posts faster than large posts) or due to filtering, or centralized numbering, or various other factors, but the result is that there may be fill-in messages delivered with sequence numbers below that of the high-water- mark, the highest sequence number seen so far. What the mark all as read functions do is mark everything below that high- water-mark as read. Traditionally, that was fine, since the number always increased, and one could be sure the messages in the next sync would all be higher numbered than the highest current message number. However, using the mark-all-read function on a server that delivers messages with out of order sequence numbers marks some read that haven't yet been seen and may yet be delivered. Thus, it's not something you want to do on servers that deliver messages out of order, if you are at all interested in seeing all messages. The alternative in simple form is when done in a group, select all messages, and use mark-read. That will only mark the ones that you actually see, and new messages should come in below the high water mark still marked unread. The problem with the simple form alternative is if you have ignored or otherwise not displayed messages. They won't get marked read. Old-pan used to have a way to setup a rule to mark such messages read automatically. Unfortunately, setting up rules was rather complex, beyond the ability of many users to do without help, and I answered quite a lot of requests for help on that over the years as a result. With new- pan Charles decided the old rules mechanism was too complex, and he never implemented it. There has been discussion of a simpler mechanism for new pan (see my reply on the New Pan Problem thread from Monday, 8 Dec 2008, 21:53:02 UTC, for some sample GUI layouts), but it hasn't yet been implemented. =:^( Thus the workaround I've been using is to set all the score-match options under View, Header pane, including view ignored. That way, they're displayed. I have the score column shown and use the color preferences to mark what's ignored, so I can mark it as read (or delete it) without even reading it. Now I can select-all headers and mark-read, and everything that pan knows about gets marked, without marking any messages yet to appear, regardless of where the high-water-mark happens to be. -- Duncan - List replies preferred. No HTML msgs. "Every nonfree program has a lord, a master -- and if you use the program, he is your master." Richard Stallman _______________________________________________ Pan-users mailing list Pan-users@nongnu.org http://lists.nongnu.org/mailman/listinfo/pan-users
