On Tue, 01 May 2007 06:46:06 +0000, Duncan wrote: > SciFi <[EMAIL PROTECTED]> posted > [EMAIL PROTECTED], excerpted below, > on Mon, 30 Apr 2007 23:24:38 +0000: > >> I also have some highly tweaked sysctl settings that help keep the >> network pipe being filled while Pan2 or Unison takes its time finishing >> the single-threaded task, then I see our network monitor sends a huge >> wad of ACKs when Pan2/Unison gets that task freed up. ;) > > Wait wait wa' wa' wait! > [...big snip, sorry...]
Duncan, have you read my bug-report I mentioned but you did not quote above? ;) I even included some screen-shots. (grrr, the gnome bugzilla is taking its time ... la-de-dah ... ;) ... ... oh here it is:) <http://bugzilla.gnome.org/show_bug.cgi?id=430628> It's very simple: When the recvspace is completely filled up (and read my bug-report for details such as how big my settings are here), no more ACKs can be sent!!! Now you can imagine how long Pan2 is taking to do things in my bug-report, even pausing the new "download thread" mechanisms, if the system is filling up those huge spaces before Pan2 has time to pull out the packets! Caution: Don't use my settings just in a whim, one must enlargen at least one other setting so the kernel / stack can grab some RAM for those huge buffers, and most kernels / stacks need to do this at boot-time (in single-user mode before the rest is primed & ready). btw I have 3.5GB of properly paired SDRAM sticks (a requirement on this model Dual G5), so I might as well use it up in areas such as this. ;) (These settings also helped when I ran the "LiftedRadio" live netradio webcasts -- yep that was "me" ;) ) I filed this problem as a bug-report so we can discuss it there and won't get lost. I have a feeling, and mentioned this in the bug-report, that the thread mechanisms aren't really "OS-native" and might be intrinsically designed that way in gtk/glib, because we have native news-readers & web-browsers that don't act this way (i.e. a FireFox download does not get "stuck" when it is busy drawing a web-page). OTOH OSX does have a problem when needing to (what I see as) refresh its font-lists (I have well over 700 typefaces installed here), it will necessarily pause the entire WindowServer (the Mac equivalent to the XServer, something coming from the NeXT days) -- but this is not what Pan2 is doing when it pauses everything under its own umbrella. This problem is really driving me crazy, and by the feedback coming from Unison & other native OSX apps having similar problems, Mac users won't put up with this pausing at all. Thanks for discussing this, but somehow we should be doing this over in the bug-report to keep it alive and to show how many other Pan2 users are affected & interested in getting it fixed. ;) _______________________________________________ Pan-users mailing list Pan-users@nongnu.org http://lists.nongnu.org/mailman/listinfo/pan-users
