Re: [Pan-users] Re: [Pan 0.133-Windows] Drops server

[Kurt Schilling]

Duncan wrote:
> Kurt Schilling <address@hidden>
> posted address@hidden, excerpted below, on  Fri, 26 Sep
> 2008 07:44:05 -0400:
> 
>> I use 0.133 on both a Windows box and my Linux box. I currently
>> subscribe to three server feeds. Two are private servers, the third is
>> Motzarella. I have noted that as I switch between subscribed newsgroups
>> on the different servers that the connexion to Motz gets dropped. The
>> list of subscribe groups is alphbetical, so by the time I get to
>> news.software.readers Pan indicates that No Tasks remain and no new
>> headers are downloaded. I can exit the program and re-establish a
>> connexion to Motz, and Pan will then fetch new headers. Any
>> sugggestions?
> 
> Pan normally does a keep-alive, sending what is effectively an "I'm still 
> here, don't disconnect" request to each server every few minutes.  
> However, some servers still timeout after some server-configured time 
> period if that's all they are getting.
> 

That's what I sort of thought. I can continue to jack-in and refresh.
This just a minor annoyance. And thank you for the quick response.




> Here, my ISP's (outsourced) news service still disconnects after a 
> certain time -- and the pan log will show an error and apparently quit 
> sending the keep-alives after that.  However, the server /should/ 
> reconnect fairly quickly if you actually start doing something on it.  I 
> know it /normally/ does here, tho there's an occasional temporary hiccup 
> and I'll say compose a reply, have it timeout in the mean time, and won't 
> get an immediate send when I try to send.  However, if I let it sit, 
> still trying to send, a couple minutes, it'll almost always eventually 
> connect and the message will be sent.  If it doesn't, there's worse 
> server-side issues.
> 
> You might try going offline (in pan) temporarily, hitting the refresh 
> button so it queues the tasks, then going online.
> 
> Other than that I don't know what to say, except to note that you can 
> check what the OS thinks exist in terms of connections using the netstat 
> command.  It should work on both MS and Linux, tho parameters may vary.  
> If the OS still sees a normal connection, that could be a problem.  If it 
> sees one that's still stuck in time-wait, that's a different problem (the 
> connection got half-closed but the last bit of the close handshake 
> sequence disappeared somewhere).

I'll see what netstat tells me about keeping connexions alive a bit later.



> 
>> One other quick question: is there a way of stripping out the added
>> carriage returns above the sig delimiter? They add extra space to a post
>> or follow up that have to be manually deleted.
> 
> I don't find it too bad here, probably because it's just habit to delete 
> the extras, now, but you're right, there's a few extra CRs in there and 
> it is (mildly) annoying.  I'm not sure why there's so many, except that I 
> think Charles may be putting them there to help to encourage proper quote/
> reply form, instead of top posting.  If the cursor and an obvious blank 
> is at the bottom...

Heh. I'm in the habit of manually stripping the extras out now. Just
looking to see if there was a workaround or fix. It is a very minor
annoyance.



> 
> The other possibility is that it's to do with the interplay between the 
> profile and the sig handling.  If you have several posting profiles (with 
> different sigs) and try switching them, you should notice that if you've 
> deleted the extra lines, they come back when you switch profiles.

Hmm, I'll see what I have in the way of sig.txt files and check that
aspect out too.


> 
> You could play around with the various sig type options.  It has been 
> awhile since I tried that and I only tried the two text options (FWIW I 
> use text file, now), but that might change the number of blank lines 
> inserted.
> 
> If I was a C++ coder, that'd have probably "itched" long enough by now to 
> cause me to scratch it, looking into creating a patch to fix it, but I'm 
> not and obviously it's not itching bad enough for those that are to have 
> bugged and filed a patch for it yet, so...
> 

Again, thanks for the response. I'll test a few things and let you know
the outcome.
With warm regards,

Kurt Schilling/Steinhauser