LSTSRV-L Archives

LISTSERV Site Administrators' Forum

LSTSRV-L

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Topic: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Zvika Bar-Deroma <AER7101@TECHNION>
Sat, 1 Apr 89 09:51:38 IST
text/plain (44 lines)
>The  problem is  that other  people request  that SIGNOFF  * (NETWIDE  be
>removed and  replaced by a short  note telling the person  in question to
>ask his  node administrator to  perform the change  for him, and  for any
>other account which might be in the  process of being deleted at the same
>time. The reason is  very simple: this costs quite a lot  of CPU time, on
>all the servers  in the network, when the guy  usually is subscribed only
>on a few of them.
>
>How do the rest  of you feel about this? I don't really  know what to do,
>apart from trying to speed up  the thing (which quite difficult given the
>speed of REXX :-( )
>
>  Eric
 
Eric,
Despite CPU consumtion, I am very much in favour and would like to be
able to "NETWIDE  QUERY" and as June suggested "NETWIDE SUSPEND"
(skip the probably wrong syntax..).
 
About the CPU and network resource usage, I have a suggestion:
Let's have your nearest server (or nearest backbone server) store the
info concerning on which lists I am; any GLOBAL * command will be
directed by the receving server to the LISTSERV supposed to handle my
requests. In that LISTSERV, I will have a record in the (new)
SUBSCRIBE DATA file, containing:
First bit of suspended or not; then my network address (I guess for most
LISTSERV subscriptions it'll be just a userid@node, although due to
domain naming and other networks, it might be much longer for some ppl.
After that will come the subscription info., contained of nothing but
list names, so my entry will look something like:
 
1AER7101@TECHNIONLSTSRV-LMAILBOOKMAIL-L  INFONETS ........etc......
 
the 1 means I'm active (not suspended), then my net-address and
all the list names...I guess the whole thing can/should be done much
more efficiently, but that's basically the idea I was thinking of.
 
What do you think about it ? May it reduce the global resource usage ?
How hard is it to implement ? Maybe LDBASE should/could be used (can it
search very long records ? -
theoretically up to number-of-lists*8+(about ~100..) )?
 
/Zvika

ATOM RSS1 RSS2