>INDEX AUTOCAT gets you the archives which are still in the Notebook=
>path, currently '97 and '96 to date. It does NOT get you the old logs
>which were separated, one year per file, still with the old log
>designations. [...]

The problem is that your compilation notebooks aren't in the Notebook=
path.  It's that simple.  I thought we mentioned that but possibly we
didn't.  LISTSERV will only index notebooks that are in the path you
define in Notebook=.  There's no way to tell it "oh, by the way, also
index these notebooks over here so they can be accessed via the web
interface".

Have you considered making something like an AUTOCAT-ARCHIVE list that
would be closed subscription, send=owners, etc., that would exist only
to provide a separate interface to the old archives?  You could then
link to that list's archive pages from the main AUTOCAT page.

WRT the search times, my understanding is that your searches will be
just as fast either way.  Quite frankly the best reason to leave the
files in small chunks rather than to consolidate them is that this
gives LISTSERV an edge when you want postings between, say, March 1992
and July 1992.  If you have one big index for all of 1992, then LISTSERV
has to load the entire index.  If you have more granularity, ie,
monthly or weekly, then LISTSERV only has to load the indexes for the
time period in question.  So (not really knowing what algorithm Eric
uses), my <guess> is that with the large yearly files you might see
a slight improvement in non-date-limited searches (where FROM and
UNTIL are left blank) but see a degradation in searches that only
need to look at a subset of the available archives.  Either way I
don't think the difference would be enough to sneeze at, but Eric
would have to confirm that.

In any case the best way to speed up a search when you have weekly
or monthly notebooks is to limit it by date.  It's like doing a
search on WINNT-L :)  If I want to search for something that's
NT 4.0 specific, I don't look all the way back to May 1995.  Similarly
if you are pretty sure that the topic you're researching was discussed
some time this year, you should at least limit your search by adding
a FROM: spec of "January 1998".  That way LISTSERV only has to load
the indexes from January 1998 on instead of having to look through
every index it has for the list.

And of course you have to remember that if LISTSERV is doing something
else (like someone else's search) when you click the "Search" button,
you're going to have to wait your turn in the queue, which can also
lead to the impression that searches are slow (when in fact the search
itself is fairly fast, you just have to wait for LISTSERV to get to
your request to process it).

Nathan