httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jim Jagielski <...@devsys.jaguNET.com>
Subject Re: MaxRequestsPerChild and MaxKeepAliveRequests
Date Sun, 21 Mar 1999 00:03:42 GMT
Rodent of Unusual Size wrote:
> 
> Jim Jagielski wrote:
> > 
> > Recall that if we have MaxRequestsPerChild really be fixed to limit
> > the max # of actual requests, we run into the case were we could
> > overrule MaxKeepAliveRequests (say, for example, MaxKeepAliveRequests
> > is set to 100, yet the child has only 25 requests "left". If we honor
> > MaxRequestsPerChild we will stop a persistant connection "prematurely."
> 
> I believe that's our privilege.  When we accept a persistent connexion,
> we inform the client of the *maximum* number of requests that can be
> accepted on the connexion.  I'm not sure HTTP requires us to guarantee
> that we will accept up to that number.
> 
> In any event, if the Connection response field on a persistent
> connexion acceptance has the max attribute dynamically set to
> min(requests-remaining,MaxKeepAliveRequests), the issue is solved.
> Right now, with MaxKeepAliveRequests set to 100, we could end up
> processing two orders of magnitude more requests than is intuitively
> obvious.  I don't consider that good.
> 

I agree that the problem is what's "expected" and what actually happens.
Let's recall that if people have had problems, then they've gone ahead
and changed MaxRequestsPerChild to what makes sense for them. If
we now change it, then we'll be affecting them by 2 orders of
magnitude. And this change is bound to effect those moving from
1.3.x to 1.3.5 than anyone else, and it's a strange change. As I've
said, if they are using it, they have most probably already changed it
to what's "right" for them, so doing this at this stage may not
make sense. IMO the alias is most likely the more transparent, esp
if we say MaxRequestsPerChild is now depreciated or does not take
into account persistant connections. :/

I'm cool if we decide to "fix" it; I'm just thinking of the user-land
impact.
-- 
===========================================================================
   Jim Jagielski   |||   jim@jaguNET.com   |||   http://www.jaguNET.com/
            "That's no ordinary rabbit... that's the most foul,
            cruel and bad-tempered rodent you ever laid eyes on"

Mime
View raw message