NCSA X Mosaic and X server crashes

marca@ncsa.uiuc.edu (Marc Andreessen)

Mail folder: WWW Talk Jan-Mar 1993 Archives
Next message: joe@athena.mit.edu: "Documentation for libWWW version 2.0"
Previous message: Joe Germuska: "problems building the httpd daemon"
In-reply-to: Mitchell N Charity: "NCSA X Mosaic and X server crashes"
References: Mitchell N Charity: "NCSA X Mosaic and X server crashes"

Date: Fri, 19 Mar 93 16:18:42 -0600
From: marca@ncsa.uiuc.edu (Marc Andreessen)
Message-id: <9303192218.AA07105@wintermute.ncsa.uiuc.edu>
To: www-talk@nxoc01.cern.ch
Subject: NCSA X Mosaic and X server crashes
In-reply-to: <9303191512.AA20922@hq.LCS.MIT.EDU>
References: <9303191512.AA20922@hq.LCS.MIT.EDU>
X-Md4-Signature: 230b5774d3d55705db1dcc977ae4442e

Mitchell N Charity writes:
> Being delighted by xmosaic, I have been doing lots of demos.
> Unfortunately, X servers are crashing.
> 
> Using the xmosaic-0.9 and -0.10 distribution binaries,
> I have encountered crashes on a variety of servers:
>    MIT X11r5 on DEC with DEC xmosaic
>    MIT X11r4 on Vax with AIX xmosaic
>    MacX      on Mac with Sun xmosaic
> 
> But, is beta code, right?
> Well, apparently mine is the only bug report on this.
> Thus the NCSA folk responded that it was the "fault" of the servers.
> 
> So, if you have a server crash, I suggest you email marca@ncsa.uiuc.edu,
> so the problem's magnitude can be guaged.
> 
> Mitchell

Let me address this is painfully honestly as I can.  As I explained to
Mr. Charity in private email, it is literally impossible for an X
client to *cause* an X server crash.  It is literally not possible.

What is happening to Mr. Charity, of course, is that the X client is
issuing some sequence of requests (Lord knows what sequence, since
he's sent me no information from which I can attempt to guess) that is
exposing bugs in the X servers he's using.

Let me repeat again: X clients cannot *cause* X server crashes, and
there is no way for an X client to know what particular sequence of
requests will expose a particular server bug, so it's not even
possible to prevent those bugs from being exposed in a particular
server by a particular client.  There is nothing the client can do to
prevent server crashes caused by server bugs -- nothing.

Thus, yes, indeed, Mr.  Charity's crashes are the "fault" of the
servers.  When I worked at a certain Large Workstation Maker with a
Three Word Name that shall remain unidentified, incoming bug reports
of the nature "client XYZ crashes the server" were routinely
classified AND FIXED as server bugs.  It is simply the way things are.
It is simply how X is designed, and implemented.

So no, don't send mail to me because your X server is crashing.  If
your email domain is 'lcs.mit.edu', as Mr. Charity's happens to be,
and you're using a stock MIT server, I suggest you walk down the hall
or across the quad and talk to the X server experts.  If your server
was supplied by your vendor, complain to your vendor.

Marc

--
Marc Andreessen
Software Development Group
National Center for Supercomputing Applications
marca@ncsa.uiuc.edu