- I've just done a PUT of PEERS NAMES to solve Marty's problem with LISTTEST.
Marty, the change file you sent was ok but it does not take effect until the
new version of the PEERS NAMES file is stored back at all the servers in the
vicinity of yours. Transferring the jobs from the bit bucket to LISTSERV is
the proper procedure until everything is fixed. :-)
- I found a bug in the DISTRIBUTE PUT function which makes it impossible to
distribute commands to servers which have disabled the DISTRIBUTE function.
Granted, those servers would NOT execute the command anyway since they have
disabled all forms of DISTRIBUTE, but "Sending files to a LISTSERV userid
is prohibited" is not precisely the error message you would expect. I have
fixed this problem in my test 1.5g version.
- The CEARN LISTSERV was playing ping-pong with [log in to unmask] I've been
able to contact the author of the latter server and I'm investigating the
problem. I am afraid that more than 300 files have been piled up on
MARTIN's reader and that there are 300 less spoolids available for RSCS :-(
It seems that RPRLSCI kept sending commands to LISTSERV from which it kept
getting replies, then replying back, etc. Since the commands were valid,
the automatic-command-war detector did not suspend the culprit and the war
kept going on... :-(
- Local commands are explained in LISTMAST MEMO. I guess the configuration of
the LOCALCMD FILE itself is no problem, the big thing is to write the
command exec itself because there is no doc about the LISTSERV routines that
you will undoubtedly need to use :-) However, LSVTELL/LSVLTELL and LSVSENDF
are pretty easy to use and I'll do my best to keep upward compatibility on
them. Add LSVKEYWD & LSVMAIL/LSVIMAIL to the list, too. The other routines
may change, though.
- I've found another problem with the CEARN server. I found myself unable to
review the stats for EARNTECH and LINKFAIL which are "Stats=
Extended,Private" -- two lists to which I am subscribed and of which I am
an owner besides... I am even a postmaster at CEARN and it DARED tell me
I'm not authorized to review the stats!!! Did any of you see this bug
occur on another server? I'm surprised because I haven't changed the STATS
code recently and it seems not to have the bug on my server. Please note
that Olivier is on vacations until the 12th of January.
- About the INCLUDE/DESCRIBE feature et al, a few points:
1) The length of the headers causes performance degradations the first
time a list is referred to, but that's all. The keywords are then
extracted from the header, compiled, and kept in GLOBALV variables.
You can't expect significant performance improvement with shorter
headers, except on the first LIST command or suchlike.
2) I originally wanted to create a STOREHDR command that would store only
the HEADER of a list. Since the header is usually the same at all the
nodes (there is no problem about mentioning yourself in the PEERS=
statement), I thought it would make list maintenance a lot easier.
However I realized that there is a problem with the "Notebook="
keyword which is NOT the same everywhere, and maybe there will be more
such keywords in the future (don't ask me -- I see no reason to claim
there will never be any other such keyword). A possible solution would
be not to mention those keywords in the common header (which is the
only reasonable thing to do besides) and have STOREHDR replace only
those lines that appear before a given "separator" line (eg *%%%%%%%%%%
%%%%%%%%%%%). The installation dependent keywords would be placed
after this separator line. Another possibility would be to specify
"blank" keywords in the common list header, eg "Notebook= .", which
would indicate that the old value of the keyword is to be placed there
at the time the list is stored. Any comment/idea?
Well, back to the 3274s... I've just realized we had a wonderful thing called
"user entry assist" on the 3274s but it was stupidly disabled... It provides
*typematic* delete key, word-delete, skip to next/previous-word, automatic
text adjustment within left and right margins (but not justification :-( ),
even when the insert mode is active, a REAL backspace key (but only in insert
mode) and a cursor position indicator. Plus programmed tabs of course but
XEDIT can do that already. And I've *eventually* found out how to disable
the "clic" key at power on :-) But there does not seem to be a way to have
the PA1 and PA2 keys work without having to be shifted, unfortunately. :-( Oh
well, I think we can live with it...
Eric
|