CNI-COPYRIGHT postings

CNI-COPYRIGHT Readers,

                    PLEASE READ THIS NOTE.

(It contains important information about the CNI-COPYRIGHT list which will affect all readers.)

You will be receiving a considerable number of postings late this evening (Friday February 21, 1997) or early tomorrow morning (Saturday February 22, 1997).

Earlier this week, the Coalition experienced some problems with our mail exchange host, and a backlog of postings developed as a result. The problem arose because the kernel on that machine was unable to open as many process IDs (PIDS) as sendmail was requesting of it. This problem resulted in a cascading failure unless Angelo or I were able to shutdown processes manually (which is darn near impossible when you can't log onto the machine -- because a login requires a PID).

This is a fairly significant problem. I've taken several steps in order to bring this situation under control in the short term:

   o I've recompiled the kernel on the machine in question after

      raising the setting for 'maxusers'.  This will allow the machine
      to manage more processes; however, it also means that the 
      amount of active memory which can be allocated to each process 
      is smaller.  So this is not a big win.

   o  I've ordered addition RAM for the machine.  Once it has arrived,
      I will recompile the kernel with a new setting for 'physmem'
      which reflects that added memory.  This will offset the problem 
      described above.      

   o  I have gone through the ListProcessor transaction logs with a 
      fine tooth comb, and have removed any address on the system which 
      which was bouncing a considerable amount of mail recently.  This 
      number includes about 150 addresses on this list (from a total 
      of around 2000 addresses).  Most of those addresses are ones which 
      are no longer valid (expired accounts -- especially former law 
      students, non-existent host addresses, etc.), and for which the 
      ListProcessor does not generate errors to the list owner/system 
      adminitrator (a small number which I've ignored and which have 
      built up over time).

      Unfortunately, I can't guarantee all of the addresses I removed 
      were invalid.  Some may simply have been poorly designated (i.e.
      using one or several forwarding addresses, using outdated 
      gateway services, etc.).  Others may be site addresses which were
      simply experiencing a "bad mail day" earlier this week.  In 
      general, I tried to remove addresses which had been bouncing 
      mail consistently for several weeks, days, months, ... but who 
      knows?  The CNI-COPYRIGHT list has been in operation long enough 
      that there were a considerable number of these types of problems.
      Enough, in fact to consume two very long days of my life this 
      week.

      Obviously, if you are reading this, your address was not affected.
      However, if any of your colleagues comment to you (in the next 
      few days or weeks) about the lack of postings on the CNI-COPYRIGHT
      list, you might suggest to them that they re-subscribe.  B^)

   o  I've reconfigured the ListProcessor which administers the list 
      to distribute each message with 50 recipients per sendmail header
      file.  Formerly, the list was configured with 30 recipients per
      header.  This increase will allow sendmail to process a smaller
      number of physical files, which will also reduce the number of 
      process IDs it requests from the system.

      CNI-COPYRIGHT has roughly 2000 addresses subscribed to the list.
      Formerly, @ 30 addresses per message, that mean the ListProcessor
      was spooling roughly 66 pieces of physical mail for each message
      processed on the list (2000 / 30 = 66.6_).  With the new 
      configuration, it only spools 40 (2000 / 50 = 40).  Formerly, 
      the ListProcessor was spooling about 1333 individual pieces of 
      mail per day (66.6 * 20 messages per day = 1333.3_).  With the 
      new configuration, I have cut that number to 800 (40 * 20 = 800).

      The bottom line (and down side) of this reconfiguration is that 
      some sites are going to experience much slower delivery times 
      for daily CNI-COPYRIGHT postings.  What you see on your end is 
      that the postings from the list are likely to be spaced farther 
      apart in your mailbox, and arrive over a longer period of the 
      day/evening.

      The problem is that any site or domain which is fairly slow at 
      receiving incoming mail (because they have a small MX host, or
      a slow WAN circuit, or because their mail system is just acting
      up on any given day, etc.) will cause all the other 49 addresses
      on a particular piece of physical mail to have the delivery of 
      their mail delayed.  There is nothing I can do about that.

Finally, I would like to inform you that the Coalition will be migrating our ListProcessor and our mail exchange host to a new machine in the near future. In actuality, I had hoped to have finished this task by this date, but as with many such technology tasks around here the implementation schedule is slightly off the target... B^(

This migration should be seamless for you, the reader. You'll continue to address the ListProcessor using the address LISTPROC[_at_]CNI.ORG and you'll continue to send mail bound for the list to the address CNI-COPYRIGHT[_at_]CNI.ORG as well.

However, a few days before this change, I'll execute a program I have written which will send you a piece of e-mail directly (not using the ListProcessor software). This piece of e-mail is designed to ferret out additional addresses on the list which are no longer valid. Invalid addresses should generate a response which will be directed back to my program and the address in question will be removed from the list prior to the list being moved to the new machine. Although many mail systems do not route such errors properly when they are faced with e-mail that contains complex mail headers (such as the CNI-COPYRIGHT list does), most will respond appropriately when faced with a very simple header that contains only one address (the sender).

You can expect my program "notice" to appear sometime around the middle of March, 1997. When it does arrive, there will be NO NEED TO REPLY TO THE NOTICE. Thanks for taking the time to read this. I'm off to Canberra for a week to attend a conference on metadata. You'll be left in the hands of my able assistant, Angelo, during the interim and we'll make the move when I return.

--

   Craig A. Summerhill, Systems Coordinator and Program Officer
   Coalition for Networked Information
   21 Dupont Circle, N.W., Washington, D.C.   20036
   Internet: craig[_at_]cni.org   AT&Tnet (202) 296-5098