Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 12 Jul 1999 18:47:16 -0700
From:      "David Schwartz" <>
To:        "Doug" <>
Cc:        <>
Subject:   RE: 3C905 versus Intel Etherexpress PRO/100?!
Message-ID:  <000001beccd1$a2a3af60$>
In-Reply-To: <>

Next in thread | Previous in thread | Raw E-Mail | Index | Archive | Help

> 	*Nod* The argument I seemed to be hearing from the poster I
> responded to was, "Why add that optimization to the code if it only buys
> us 2%?"

	A 2% optimization that makes things more complicated would probably be a
bad tradeoff on a server or server OS. Cleanliness of implementation is a
major priority there, because you really do need rock-solid reliability and
maintainability. That tends to stress straightforward design and punish
excessive cleverness.

	I don't particularly care if I have to reboot my desktop. But it really
irritates me if I have to reboot my servers. It irritates me more if they
reboot themselves.

	However, I have to admit that while I understand this and agree with it, I
make 'dangerous' optimizations to server code all the time. All of those 2%
optimizations eventually add up to 20% optimizations -- and that does
matter. And after enough time and testing, the new and dangerous code
becomes less new and, hopefully, less dangerous.

	If I were comparing two operating systems for use as a web server, would I
care about a 15% performance difference? Would that mean anywhere near as
much the difference between 99.9 percent uptime and 99.95 percent?


To Unsubscribe: send mail to
with "unsubscribe freebsd-chat" in the body of the message

Want to link to this message? Use this URL: <$a2a3af60$021d85d1>