Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 01 Aug 1999 01:19:37 -0500
From:      Jon Hamilton <hamilton@pobox.com>
To:        Doug <Doug@gorean.org>
Cc:        Sheldon Hearn <sheldonh@uunet.co.za>, Matthew Dillon <dillon@apollo.backplane.com>, hackers@FreeBSD.ORG
Subject:   Re: Mentioning RFC numbers in /etc/services 
Message-ID:  <19990801061937.874C5135@woodstock.monkey.net>
In-Reply-To: Your message of "Sat, 31 Jul 1999 22:58:27 PDT." <37A3E203.DE0FE656@gorean.org> 

next in thread | previous in thread | raw e-mail | index | archive | help

In message <37A3E203.DE0FE656@gorean.org>, Doug wrote:
} Sheldon Hearn wrote:
} > 
} > On Fri, 30 Jul 1999 15:05:14 MST, Doug wrote:
} > 
} > >       I still haven't heard anyone answer the two key (IMO) questions.
} > 
} > Your questions are easier answered in reverse order:
} > 
} > > and how do you justify the additional cost to parse the file for every
} > > single system call that uses it?
} > 
} > The information is part of the comments within the file. The cost is in
} > disk space, and it's cheaper than fleabites.
} 
} 	Nowhere did I mention disk space. I agree that if that were the only is
} sue
} I wouldn't be raising the objection. 
} 
} > > Why is it better to have this in the file than in a man page,
} > 
} > Since it costs nothing to have it in /etc/services, why not leave it
} > there along with the information with which it's associated? The
} > alternative is to have a manpage that contains most of the information
} > in /etc/services!
} 
} 	And why is that bad? Since when is redundancy in the documentation a
} problem? Like you said, disk is cheap. 
}  
} > > My only concern is that putting it IN the file has more costs than
} > > benefits.
} > 
} > What am I missing here, that I don't see a cost? What software scans the
} > lines in /etc/services beyond the comment delimiter, other than null
} > terminator searches?
} 
} 	So, how many comments are you going to add? Let's say the total parsing
} cost to the system for all of your additions is X. X is probably a pretty
} small number, I'm not arguing that point at all. Now multiply X times every
} packet on a highly loaded server, because that's how many times ipfw is
} going to need to parse the file (roughly). 

No.  ipfw deals with /etc/services only at startup time (any other behavior on 
its part would be ridiculous).

} 	My point is simply that the information is valuable, but it belongs in 
} a
} man page. There is no reason to add a good deal of non-functional
} information to a file that is used by so many parts of the system. 

I think you're  overestimating the cost by a considerable amount.  I'm 
not saying that the cost is zero, but I am saying that it's close enough 
that the value of having the information *right there* outweighs the cost.
Can you demonstrate a realistic scenario in which multiplying the volume
of comments in /etc/services by, say, 10x results in a perceptible 
performance hit?

-- 
   Jon Hamilton  
   hamilton@pobox.com



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990801061937.874C5135>