Re: [Pan-users] Proper behavior of "backup" servers? [DIAGNOSED]

[Duncan]

Heinrich Müller posted on Wed, 23 Nov 2011 10:27:17 +0000 as excerpted:

> Am Tue, 22 Nov 2011 19:51:18 +0000 schrieb walt:
> 
>> On Sun, 20 Nov 2011 16:29:41 +0000, walt wrote:
>> 
>>> I discovered that setting gmane.org to "backup server" prevents pan
>>> from ever asking gmane for articles that I click on in the header
>>> pane.
>> 
>> The reason pan was refusing to fetch articles from gmane.org is that I
>> had another server set as "primary" AND set to zero connections.
>> 
>> I think pan was waiting for that primary server to ask for the article
>> first -- but of course pan couldn't ask the primary server because its
>> number of connections was set to zero.  Setting that primary server to
>> "fallback" allowed pan to query all the other fallback servers,
>> including gmane.
>> 
>> This isn't exactly a bug, but the behavior is unexpected.  Maybe pan
>> should refuse to set a primary server to zero connections?  Or at least
>> give us a warning?
> 
> I'll implement 1 as the minimum for the spinner, good idea.

FWIW, if you check the history, "1" was the original minimum number of 
connections in Charles' code.  Someone asked for and got an enhancement 
that allowed disabling a server by setting 0 connections.

I /think/ that enhancement was added by khaley, post-charles, but it 
/might/ have been one of the last commits Charles made.

Meanwhile, I actually use the 0-connections=disabled-server feature here, 
as I've kept around several years of posts from my ISP's former server, 
now long decommissioned.  If you remove the 0-connections=disabled 
feature, please add an explicit enable/disable checkbox or the like in 
its place.

Alternatively, leave 0-connections=disabled as-is, but change the logic 
of the primary/backup/multilevel server system so that disabled servers 
are skipped, basically treating them as if the post wasn't found on the 
disabled server at all (which it wasn't, because the server was 
disabled).  This would be my preferred outcome, and I don't /believe/ it 
should be too difficult to implement (simply early-out with a nack on the 
check if a message was available on a server, if it's disabled), but I 
don't claim to be a coder and haven't looked at the code to see how 
difficult in practice it might 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