Iain Shigeoka wrote:
>
> I'm concerned about the size of our wiki from a strictly performance
> and organization standpoint. From my limited unix knowledge I
> recall about 60 files per directory being a nice limit to keep
> directory speed optimal. Large directories shouldn't get more than
> a few hundred files in them or performance begins to significantly
> degrade for all file system operations.
>
> At last count, the joswiki has well over 1000 wiki pages (I believe it
> was 1200 or 1300 last time I checked). I've been noticing some
> pretty poor performance as a result.

We have closed to 800 pages in our internal TWiki.Know
knowledge base web (a total of 1600 files with the RCS
included). I don't see any performance penalty compared
to other TWiki webs on the same server with fewer
pages (50 or so). I did not measure it, however what I
can tell from using the system it seems like Unix can
handle lots of files in one directory very efficiently.

> BTW, for those with this size site, you need to be careful in using
> the search feature of the Twiki. Grepping into such fat directories
> will absolutely kill your webserver.

Grep search, as you point out, does get slower also on
our system. Still, the speed is acceptable for us. We
are running TWiki on a Linux box. Probably also a
question of hardware. Don't ask me what we have, the
machine sits in Austria, I am in California. (We are
more concerned about the delay and unpredictable access
of pages across an ocean.)

The latest Beta of TWiki has the search part externalized
into a separate file, wikisearch.pm. It should be straight
forward to replace that file with an interface to a search
engine like ht://Dig ( http://www.htdig.org/ ) or
Webglimpse ( http://webglimpse.net ).

> I would also be interested in any other ideas for automating self-
> healing and self-organizing a wiki.

A page delete and rename function would help to
get rid of unwanted and orphaned pages.

Bob's Kehei Wiki has that implemented nicely,
and protected with author rights as he pointed
out in his reply.

I am considering a page delete and rename function
without author rights. This could be dangerous,
so the user interface must be designed carefully.
How about this: When you delete or rename a page,
a delete / rename confirmation page would show a
list of all pages that have a link to the page to
be deleted / renamed, and a recommendation to update
those pages first before proceeding. (The rename
function could do that automatically). Deleted
pages go to a Trash Can web. An undo delete
function could restore pages that have been deleted
unintentionally.

-- PeterThoeny - 08 Mar 2000
-- http://www.mindspring.com/~peterthoeny/twiki/