Re: [Pan-users] cache NNTPMessage from localhost

[Duncan]

Thufir posted on Wed, 04 Apr 2012 08:49:02 -0700 as excerpted:

> I'm playing around with gnu.mail.providers.nntp.NNTPMessage and it
> behaves in an unanticipated way -- but as documented.
> 
> If the store is closed this invalidates the messages, so the NNTPMessage
> is only useful when connected to the server.  I'm only connecting to
> leafnode on localhost.

I'm guessing that's a gnu nntp library function?  I'm not a dev and don't 
have a clue on that.

> Without diving too deeply into the workings of Pan, how does Pan handle
> this stuff?  Pan copies each message into pan2/article-cache/ using its
> Message-ID as the name for a unique ID, ok.
> 
> Hypothetically, could Pan (or my reader) work off /var/spool/news/
> message.id/ instead, which is where Leafnode caches its articles?  Not
> that there's a real point to that, just curious.

Your reader can do what it wants...

Pan... if the message-id conversion to filenames is the same as 
leafnode's, it should in theory work, but be sure character-conversion is 
the same for the corner-cases.

Also, note that pan's cache size defaults to 10 MB.  Presumably if you're 
working directly off the leafnode cache, you'll want to set pan's cache 
size large enough that pan never deletes anything, so leafnode can 
continue to manage it.  Either that or do the reverse, let pan control 
the cache size and set leafnode's far larger.  FWIW, I've had mine set to 
several gigs on both my binary and text pan instance, for years, keeping 
a more or less permanent archive of text groups in that instance, and 
managing the binary instance (back when I did binaries, I've not done 
anything with them but for the occasional attachment in a text/technical 
group/list, in years) manually.  So I know that pan's cache size setting 
expands without issue into the 10s of gigabytes, anyway.

> Another direction is to just keep the socket to Leafnode open, and then
> if it gets closed inadvertently, just open a new connection.  Does that
> seem at all reasonable?  I would prefer to deal with individual messages
> only at the level of NNTPMessage.

I still don't know anything beyond what's in your message about 
NNTPMessage, but the idea in general appears to me to be sound.  Long 
term socket connections, ESPECIALLY localhost only, should "just work".  
If they don't, there's something else seriously wrong.



-- 
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