Corporate Technical Memory Architecture (was: Re: Lotus Notes -- Too

Fisher Mark (FisherM@is3.indy.tce.com)
Wed, 14 Sep 94 06:45:00 PDT


Here, due to overwhelming demand (or at least whelming demand :)...

The TCE Corporate Technical Memory (CTM) is a reference document repository
constructed from World-Wide Web clients and servers, CGI scripts, WAIS, a
short Visual C++ program, Microsoft's protected-mode TCP/IP for Windows, and
a document submission/maintenance daemon written in Perl.

The document submission scenario (the part most imperfectly handled by
current WWW clients) is:
1) The user specifies a document title, their name, a document file, and
optionally a document abstract (abstracts are forced to be written in HTML
so as to provide references).
2) This data is placed into an ASCII control file (similar to the old UNIX
lp daemon control files).
3) The control file and data files are FTP'ed to a spool directory on the
CTM server. Microsoft's protected-mode TCP/IP for Windows has a
command-file interface like UNIX FTP, making it easy to construct a command
file to do this rather than having to code FTP directly.
4) The CTM daemon empirically determines whether the document file is a text
file, a Word for Windows file (our corporate WP standard), or a binary file.
5) The document file (and abstract file) are placed in a standard location
with each author having their own set of files. Older versions (if there
are any) are also placed in a standard location.
6) The document file is indexed via WAIS. 3 items of note:
a) Word for Windows documents are indexed by their
document text.
b) If a document abstract exists for a binary file, that
abstract is indexed in place of the binary file.
c) The daemon re-indexes all documents nightly, as the
updating of a document creates two references to it
in freeWAIS.
7) Author and title databases are updated.

Documents can be browsed by title, by abstract for title, and by author.
Documents can be searched by WAIS full-text search, by author, by title,
and by abstract for title.

The control-file method for submitting files is old technology, but it is
inherently serialized (one daemon handles one file set at a time) and is
still suitable for applications where real-time response is not required.

The submission program could instead be implemented as a WWW form if a
client had a "submit this whole file" widget. I envision such a widget
looking something like:

+---------------------------+ /---------\
Document: | | | Open... |
+---------------------------+ \---------/
where the "Open..." is a labeled button that brings up a dialog box allowing
you to select a local file rather than typing the name directly into the
text box.

BTW, any pointers to where a paper on this might be submitted would be
welcomed...
======================================================================
Mark Fisher Thomson Consumer Electronics
fisherm@indy.tce.com Indianapolis, IN

"Just as you should not underestimate the bandwidth of a station wagon
traveling 65 mph filled with 8mm tapes, you should not overestimate
the bandwidth of FTP by mail."