Sender: |
The Revised LISTSERV Distribution List <LSTSRV-L@EB0UB011> |
Subject: |
|
From: |
Harri Salminen <LK-HS@FINHUTC> |
Date: |
Mon, 04 May 87 14:35:39 FIN |
In-Reply-To: |
Message of Fri, 1 May 1987 18:06 SET from <ERIC@FRECP11> |
Reply-To: |
The Revised LISTSERV Distribution List <LSTSRV-L@EB0UB011> |
Parts/Attachments: |
|
|
Eric,
you can count us in the backbone and update our server automatically. We
also try to autolog the server every hour in case there isn't any
persistingproblems like disk full when listserv should shutdown immediately
and stay so. I guess that only vacations might cause delays, if both I and
Petri decide to take it at the same time :-).
One thing that concerns me with these peered filelists and updates is what
happens if one postmaster at the other end of a large peer network updates
peered files before he has received the latest version from the other end due
delays. The locking of files isn't feasible because it would triple the time
needed for update in the worst case. I think a simple method for detecting
this hopefully rare incident would be enough:
A) Keep an automatically assigned version number with each peer file. When a
file with a version number different than current + 1 arrives both are held
or bumbed to postmaster and a warning sent to the originator and the
postmaster of the problem site to manually correct the situation.
B) Try to send an advance warning of the update to the postmasters of all
servers you are trying to update using messages. This could be best done
with a relay style interserver message protocoll e.g. a special command
which sends a single line freeform message to all postmasters of the
server.
This feature might be usefull also for real people who are trying to inform
postmasters of some particular node of some urgent situation like a loop.
This feature need not be 100% reliable and could be turned off selectively
but it would help getting postmasters rapidly informed by remote users and
servers not knowing all postmasters of a given node (do you still remember
all the hidden postmasters you might have?). "ALERTOP message" might be the
hypersimple syntax but the special peerupdate warnings might be trapped and
shown only if they concern that server in the same manner you'll discard
unnecessary PUT commands.
Hopefully we'll have a multiwindow (might be real fun with VM/SP rel. 5 :-))
syntax oriented listeditor to manage those files without having to purchase
the VM/Prolog :-)
Harri
|
|
|