I've almost finished to change the 250+ lists on which I was an owner to have
MABOGEN@DEARN instead, including some lists which still pointed to MEYER@DEARN
due to my lazyness :-) This is almost finished (I'm waiting for some acks,
have to coordinate with Thomas two lists which he was updating at the same
time, and there will still be about 4 pathological cases). It took less than
three days. Since peered lists are reputed (not without reason) to be very
boring and difficult to maintain, I've thought I should share the methods I've
used.
* I've found invaluable to have a fast, local LISTS database (fast is here
much more important than local) to see of which lists I was an owner. I
really didn't remember. I found it also very useful the fact that the LISTS
database is always more-or-less up-to-date: if I changed some lists, next
day they didn't appear any more as results of my queries.
* I've used LSVTALK all the time, and I found it very useful :-)
* The peering method for the BITNIC lists has proven to be very effective
(I'll talk more about that later).
I've also found some problems, tho: in some lists, the set of owners is
completely inconsistent (in some cases, the intersection is the empty set,
which leaves us without any global owner); and even the passwords sometimes do
not correspond. Anyway for me the worse is how inadequate and/or outdated the
list descriptions are, if they exist at all.
It's also very annoying to find tons of purely local lists in the LISTS
database. Once again: PLEASE change all your lists which you don't intend to
be network-wide to
Confidential= Service Service= something
where something could well be 'Local' in many cases. And please, please: do
care about what you put in the list. It's not funny to look at a list and find
that it's description is "Dummy comment line" ;-)
Since the schema I've been using for the BITNIC lists works so well, I think
most lists would benefit if they were migrated to this method. I guess all
this didn't get enough audience because it was posted to REDIST-L only. If
there is enough interest, I can try to find the original posting and resend it
here.
What I'm proposing is the following: let's take REXXLIST as an example.
REXXLIST has peers at DEARN, UGA, UCF1VM, FINHUTC, POLYGRAF, EB0UB011, HEARN
and OHSTVMA. The postmasters for all these nodes should create a REXXLIST
FILELIST (this is a one-time only job), peered between these nodes. This
filelist could later be used for other purposes (for example to collect some
important files), but for our discussion they should contain at least the
following:
* A file called REXXLIST KEYWORDS, containing the network-wide keywords,
global to all peers, *including an owner for every peer*.
* A *local* file called REXXLCL KEYWORDS, containing peer-dependent
information (maybe the Notebook= option, or Errors-To).
* Another file called REXXLIST DESCRIPT, with a decent description of the
purpose of the list, methods to subscribe, associated files, etc.
* Another file called REXXLIST TOPOLOGY, containing a nice topology map for
the list.
At all peers, the list should then look as follows:
* The VM/SP REXX Language Discussion List
*.IK
*.IK REXXLCL
*.IT
*.IT REXXLIST TOPOLOGY
(password and subscribers
Now what are the advantages of that (since until now it only appears to be
much more complicated)?
- If an owner changes, you simply edit REXXLIST KEYWORDS, change the owner,
and LSVPUT the file to your server (and wait for the acks :-)). The file
automatically propagates to all peers, and you've not to use this boring GET
+ PUT process.
- If you do a topology change, you should no longer be lazy to update all
peers just for that, since a simple LSVPUT of REXXLIST TOPOLOGY will do it
at all peers.
- The same is true for the list description, in case for some reason it has to
be changed.
- Etc.
This was proposed a long time ago by Eric, in a slightly different way, for
PEERS FILELIST. But PEERS FILELIST seems to have been abandoned, and in the
meanwhile this could make the life of most postmasters much easier,
homogeneize the lists, which is a Good Thing, and, as a side effect, give to a
lot of people some familiarity with FILELISTs, which I think also is a Good
Thing... :-)
Jose Maria
|