Warning: the reply address points to me, not the list.
I would like to start beta-testing 1.7d either by the middle of this week
(after the Ebone meetings) or 1 week later, depending on resulting
workload. Please apply now if you are interested. I need mostly SP5 sites
and XA/5.5 sites, since I am running ESA 1.1 myself.
The differences are mostly internal changes which to improve robustness,
functionality and performance while decreasing dependence on CMS
functions such as the various reader file manipulation commands, XEDIT
sort, and so on. The REXX internals were mostly approximative
implementations of the syntax they described - for instance, the QUIET:
keyword seldom worked as expected when the syntax was not a straight list
of userid@node's, trailing .BITNET's were not consistently removed, and
so on. REXX code seldom displayed error messages when system commands
gave an unexpected return code, and these messages were usually not very
easy to understand anyway. Things are better under 1.7d, but there are
still 25k lines of REXX and thousands of potential errors that come from
CMS and for which one cannot obtain a meaningful description.
The main performance improvements are for LIST GLOBAL, REVIEW, processing
of incoming Netdata/DISK DUMP files, generation of BITEARN NODES and
PEERS NAMES-based tables, maintenance of large lists, and when mailing
files to Internet users (in particular LSVMAIL should not run out of
storage any longer - not that any recent change had made LSVMAIL require
more storage, but for some reason many people think it is preposterous of
me to suggest adding 1M of storage when moving from CMS5/1.7a to
CMS8/1.7c and that the reason I make LISTSERV issue warnings at startup
when it thinks it hasn't got enough storage is so I can be notified when
people reboot their servers). LISTSERV will need more storage when doing
nothing due to the fact that REXX code which is not in use sits on disk
while PASCAL routines are always in memory, if you are not prepared to
add 200k if LISTSERV complains it hasn't got enough, stay with 1.7c. On
the other hand, many things that required lots of memory now use up much
less, due to not having to load things into stems, edit them, then stack
the result and call a routine which will pull the stuff back to another
stem and finally stack it again for some other command.
The main functionality improvements are new REVIEW options to sort the
subscribers' list, better error reporting, automatic conversion of
DISTRIBUTE jobs in Netdata or other recognized format before delivery to
mail-only recipients, support of ':oldnode.'/':newnode.' tags, automatic
recognition of local Internet address if listed in BITEARN NODES but not
in configuration file (this seems trivial but can avoid many mailing
loops), 'Reply-To:' field pointing to list owners on most list management
messages sent to end users, QUERY * for list owners, and many small
improvements to the SHOW command.
The main incompatibilities were already posted to this list as the code
was being written. There is a last one I didn't mention yet - LSVBESTP is
gone completely. If you have local programs which called LSVBESTP with 4
or more parameters, you will have problems (the 4th parameter was blindly
INTERPRETed, which the new PASCAL code cannot quite do). If you only call
it with 2 or 3 parameters, you will be able to use LSVBSP with the same
calling sequence. The name was changed to make sure all attempts to call
the old routine are caught.
Eric
|