Hi Sam,

On Thu, Dec 20, 2001 at 05:39:16PM -0500, Sam Varshavchik wrote:
 > Stephane Enten writes: 
 > 
 > >On Fri, Dec 07, 2001 at 03:35:37PM +0100, Olivier Poitrey wrote:
 > > > Hello,
 > > > 
 > > > Authdaemon system don't seem to have any cache system. For authmethod 
 > > that
 > > > need database access, I think that it's not really powerfull. What about 
 > > > implementing a cache mechanism in authdaemon ?
 > > > 
 > > > I'm sorry if it is already in the TODO list. 
 > >
 > >What about this unanswered request ? 
 > >
 > >I think many of us using some database (ie. MySQL or LDAP) would really
 > >like to have authdaemond make use of the fact that he's persistant.
 > 
 > It already does.  It keeps a persistent database connection open. 

And he's nice to do so.
If he didn't I hope you would probably had make it a library instead.

 > >When it comes to tens or hundreds of queries per second, having some
 > >caching would save us some serious load on the database server(s).
 > 
 > Before you design these kinds of grandiose schemes, try to do some analysis 
 > first.  Caching will only work if you get the same requests, over and over 
 > again.  Unless you have lusers that try to log in hundreds of times per 
 > second, caching only adds unnecessary complexity, opportunities for more 
 > bugs to fester around, and you won't end up saving anything since nearly 
 > every request won't be found in the cache, and you end up going to the DB 
 > anyway. 

Well, in the real world I meet daily, users are checking their mails
every few seconds or minutes, my servers have plenty of RAM and I would 
hope that using this RAM to cache, even a *lot* of authentication
entries could help my database servers to life more idle-y.

That said, I understand your point of view ... are your against the DNS
caching concept too ?

 > So, exactly what do you expect to get from caching, really? 

Well, given a million of users (or more) in the DB, and that a majority
of users are living their mails as a session (ie. they start
reading/answering their mails, then stop and do something else, or even
disconnect if they are on dialup), I guess that some windows are created,
and that a cache could cope with that.

 > >Another great benefit of a caching engine would be that DoS'ing the
 > >database by hammering the services with authentification requests would
 > >become a lot more difficult.
 > 
 > There are other things in both the IMAP, POP3, ESMTP, and Webmail servers 
 > that are designed to prevent these kinds of attacks. 

Are you talking about the MAXPERIP option that we can't omit without
seeing -maxperip complaining about the lack of value ? :)
(tried to set as a value of 0, complained the same way)

Sadly enouht, I can't use that since I can't set them at a reasonable
value unless I have my customers behind some NAT (and belive me hundreds
of employees NAT'd as a single IP isn't that unusuall) wanting to burn
my house.

Regards,

        Stephane


_______________________________________________
courier-users mailing list
[EMAIL PROTECTED]
Unsubscribe: https://lists.sourceforge.net/lists/listinfo/courier-users

Reply via email to