Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!think.com!yale!cs.yale.edu!yarvin-norman
From: yarvin-norman@cs.yale.edu (Norman Yarvin)
Newsgroups: comp.unix.wizards
Subject: Re: Ware Ware Wizardjin
Message-ID: <29992@cs.yale.edu>
Date: 12 Apr 91 03:50:20 GMT
References: <9104072151.AA28702@gaia> <1991Apr9.020525.13001@mtxinu.COM>
Sender: news@cs.yale.edu
Lines: 25
Nntp-Posting-Host: turquoise.systemsx.cs.yale.edu
Originator: yarvin@turquoise.CS.Yale.Edu

ed@mtxinu.COM (Ed Gould) writes:
>  There is no
>good technical reason that a small, efficient X server couldn't be
>written.

To initialize a connection and open a window with the X protocol takes 12
bytes sent from client to server, followed by ~130 bytes (minimum) sent from
server to client, followed by 40 bytes (minimum) from client to server then
32 bytes from server to client.

Pure communication time is possibly the least of the overheads which this
profligacy generates.  Time gets spent generating and interpreting this
stream.  Since programmers would hate to do this themselves, a library must
be written to keep them happy, adding another layer of overhead.  Since the
raw byte stream is nearly unusable the library has to be completely general,
but that makes it tedious to use; this leads to other libraries on top of the
basic one.  These libraries add overhead both in time and in memory; if
shared libraries are not used the latter cost becomes nearly prohibitive.
All these seem to me to largely be consequences of the design of the
protocol.

--
Norman Yarvin					yarvin-norman@cs.yale.edu
 "We are at the moment in such a deep state of confusion that we are bound to
 learn something from it."  -- Sir Rudolph Peirls