Re: Cache and agent ?

[Kevin Greiner <kevin.greiner@vignette.com>]

Xavier Maillard <zedek@gnu-rox.org> writes:

> On 21 sep 2004, Kevin Greiner wrote:
>
>> Xavier Maillard <zedek@gnu-rox.org> writes:
>> 
>> > Hello,
>> > 
>> > Is it non-sense to have articles cached and have it
>> > downloaded by the agent at the same time ?
>> > 
>> > How is it processed internally then ?
>> > 
>> > (I have a group full of downloaded *and* cached articles due
>> > to a problem on my side).
>> 
>> The agent and cache are completely independent. So you have two
>> copies of each article on your harddisk.
>
> So, my next question is simply: is this situation safe ? Err,
> from where Gnus fetch the article then ? (cache ?, agent ?,
> randomly ?).

It's deterministic; it should prefer the cache.  The actual logic is
recursive and, at times, obscure so some functions may check the agent
first.

  
>> In anticipation of the next question, the agent and cache were
>> designed this way deliberately.
>
> No problem with how this was done ;) just a question. Do you
> think caching adn agentizing in the same time, a recommended/safe
> practice ? Oh, and given the fact I am caching ticked/dormant and
> even unread articles, can I safely enable agent expiration on my
> groups ? If I understand correctly, caching prevent me from a
> server expiration and adjusting in this case 2 or 3 more
> variables I can still keep a copy of some articles locally (no
> expiration can apply).

Very good, you found the primary justification for keeping the cache
and agent separate.  Agent expiration only applies to the agent's
folders so your cached articles can NOT be expired.  Note: An article
that is both cached and agentized may, by expiration, end up only
cached.

I'm not sure about your 'server expiration' comment.  My interest has
largely been the agent.

Kevin