Subject: Re: CVS commit: syssrc/sys/kern
To: matthew green <mrg@eterna.com.au>
From: Perry E. Metzger <perry@piermont.com>
List: tech-kern
Date: 12/06/2002 19:48:19
matthew green <mrg@eterna.com.au> writes:
>    You mean like, perhaps, a system already on the edge of collapse where
>    the delays will make the collapse worse, as has already been explained
>    by others?
> 
> so, if fork() is failing how is the system actually going to get the
> work done it needs to get done anyway?  that's the bit i don't get.

thttpd would only fork if it tried to run a cgi. Static content could
still be served without any trouble though. Since thttpd is event
driven (which is why it kicks apache's ass in terms of performance),
the pause stops ALL i/o, not just that for the cgi requester. Bad bad
bad bad bad.

It is true that you could hack thttpd to avoid this happening, but it
is only the easiest example of why the idea is bad, it is hardly the
only one. In general, I don't want my processes to block unless they
ask to, period -- it screws up everything from userland threads
packages to all event driven servers.

If it is to be kept, make it an option and don't set it by
default. Then I don't care -- other people can do whatever damage they
like to their own machines. Leave mine alone....


-- 
Perry E. Metzger		perry@piermont.com