<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
</head>
<body>
<p>The octal code for the <font size="+1"><tt>DECnet Frame Too Long</tt></font>
error is <u>240</u>. In order to free up some PS: space on my
systems, I split off all these off into a separate binary file on
another structure. On <font size="+1"><tt>VENTI2::</tt></font>,
<font size="+1"><tt>ERROR.SYS</tt></font> went from 96,967 pages
to 45,847 pages. The separate binary file containing only the
240's is 51,120 pages long, so more than half the space were these
errors.</p>
<p><font size="+1"><tt>TOMMYT::</tt></font> was a similar story:
there, the <tt><font size="+1">ERROR.SYS</font></tt> file was a
whopping 137,970 pages long. With the 240 items split off, it
went down to 9,614 pages, while the (binary) extracted items were
129,410 pages long. It's hard to understand what files of these
sizes mean; at nearly two RP06's, this single file approaching ⅔
of the total storage that was allocated to ⅓ of Columbia's 25,000
undergraduate student body.<br>
</p>
<p>The size of my ERROR.SYS files is due to two factors:</p>
<ol>
<li>It's every error since I started running Tops-20 again in late
2001</li>
<li>I have a lot of errors due to DECnet nodes coming on and off
line</li>
</ol>
<p>The 2<sup>nd</sup> item is triggered because of a nightly batch
job (<font size="+1"><tt>VIKING</tt></font>) that updates my
development changes on <font size="+1"><tt>VENTI2</tt></font> to
another structure on <font size="+1"><tt>TOMMYT</tt></font>, in
case I blow <font size="+1"><tt>VENTI2</tt></font> up. The KLH10
NI can pump out an amazing amount of data, but chokes when it is
receiving; I've never really understood why (not that I've
carefully looked). If you want to get data into the machines,
then you either have to use a magnetic tape or Kermit.<br>
</p>
<p>Anyway, this plus, quarterly backups, is how I address losing
files by mistake. You ARE all doing backups, right?</p>
<p>Meanwhile, it seems to me that, for the time being, I really
don't care about the node-online/offline events. Unfortunately,
there does not appear to be any way to shut them off. I'm
thinking putting together some kind of monthly batch job to strip
these out. Again, any Tops-10 or Tops-20 machine is going to see
<i>huge</i> error logs because of what's happening.</p>
<p>Isn't anybody else noticing anything?</p>
<p><br>
</p>
<p><br>
</p>
</body>
</html>