Re: [leafnode-list] groups listed twice in groupinfo, and other

[Matthias Andree <ma@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>]

Joerg Dietrich <joerg@xxxxxxxxxxxx> writes:

> On Sun, Nov 10, 2002 at 06:59:22PM +0100, Matthias Andree wrote:
>> How can I do that and still meed these two criteria without caching each
>> server's group list?
>> 
>> 1. groups that are not on *any* upstream any more must go away
>> 
>> 2. groups that are on that particular failing upstream server must not
>>    lose their locally-stored information such as article pointers,
>>    description, counters and so on.
>
> Internally, e.g. in groupinfo, store the information from which
> upstream server(s) a group is received.

I fear that is even more expensive than keeping separate listings.

I've also considered this (not for 1.9) for fetchnews.

Define two data structures, group_present, group_missing, as set or hash
or something with quick random access.

1. If a group is found, put it into the "group_present" store.
   If it is in the "group_missing" store, erase it there.

2. If a group is not found (411), then if it is not in the
   "group_present" store, put it into the "group_missing" store.

3. At the end, kill all groups from active when we were able to talk to
   all servers.

> Not really cheaper than caching. The feature request is something I
> considered more than once and is really worth having, IMHO.

Not in 1.9.x though.

-- 
Matthias Andree

http://sourceforge.net/projects/leafnode/   - leafnode-1 tarballs
http://mandree.home.pages.de/leafnode/beta/ - leafnode-2 tarballs

-- 
leafnode-list@xxxxxxxxxxxxxxxxxxxxxxxxxxxx -- mailing list for leafnode
To unsubscribe, send mail with "unsubscribe" in the subject to the list