Timothy, others -

Seems there's enough interest to warrant posting to the list.

 > Has anyone implemented EAD 1.0 with OpenText Livelink?
 > If so, I would appreciate the chance to exchange notes on
 > how to set up the software.

We use Open Text (OT) Livelink 7 for indexing & retrieval (I/R) of our
cross-institutional DB of around 5000 Finding Aids. Some are beta, some
version 1. We haven't seen any problems with version 1, but then we
don't use OT's presentation facilities directly.

One reason we chose OT in the first place was their sgml
support.  This seems to be going away, & never worked properly for us
anyhow.  Their sgml processor required that all tags be explicitly
closed.  Since we had no control over the encoding of the documents, we
couldn't conform with this requirement. Instead, we use their 'tagged
ascii' indexing feature, which seems to work about like the sgml would
have, allowing us to use tag names as fields in searches. I think the
only thing we lose by doing 'tagged ascii' rather than sgml is the
ability to use attribute names as field names in searches. This doesn't
seem to have much practical impact for us.

A more significant shortcoming of OT's indexing, for us, is the lack of
a structural notation in fielded searches: i.e., if you have an element
X that can occur withing element A & element B, there's apparently no
way to say 'consider only X values within A'. I'm not sure if this is
because we're indexing as 'tagged ascii' - maybe you can do this with
their sgml indexing? I'd like to hear about reasonably priced I/R
software that will do this, for sgml & xml. I think verity will, but
it's out of our price range.

We like OT's I/R power very much The snappy performance compensates
pretty well for the middleware we encumber it with. We came up with the
middleware because we didn't care at all for their out-of-the-box user
interface. There is a command-line interface to their search software,
but it's undocumented & unsupported. Some of the pat50 commands
documented by michigan (?) seem to work the same way with the 'pat60'
that came with our bundle, but we couldn't figure out how to use it in
the cgi environment to do things like deliver a stream of summary-level
elements from a result set. (I can dig other complaints/puzzles out of
my notes if anyone's interested.) So we ended up interposing a web
interface of our own devising between the users & the OT web server.
This middleware gives us the look we want, at some performance cost - at
a very minimum, it doubles the hit rate, since each post to our
middleware becomes at least 1 more post to the livelink httpd, &
sometimes more.

Originally we thought web browsers would support sgml more widely than
has turned out to be the case.  Many folks found Panorama unpleasant
though, & nothing better came along, so we decided to serve html. We
were disappointed with OT's html conversion, & contracted with UC's Alvin
Pollock for some enhancements to his ead2html proofing tool, to make it
more general & configurable. We are quite pleased with the quality of
the html produced by ead2html, & with its flexibility. It isn't fast
enough for on-the-fly conversions, though, so we make static conversions
as part of the DB maintenance cycle, & serve these.

Another reason we chose OT was that they have an integrated web crawler,
which we thought would be a nifty way to do data acquisition. While
it may have been nifty, it has not turned out to have much operational
value for us, because of crawler limitations & variations in the ways
our suppliers configure their sites.

We looked at Dynatext|web at the same time (mid '96) we were evaluating
OT, but felt they were too slow & also didn't care for their html
view. Currently their html looks much better, e.g. the way Duke has
tweaked it.

We don't really think we fit OT's customer profile any more, if we ever
did, & this is one reason we're looking to replace them.  They don't
offer us educational discounts (nor do INSO) - no doubt we'd find them
more likeable if they did.  We're thinking SWISH or Thunderstone may
work for us, & would be interested in any comments from folks who have
used these on sgml DBs. I'd be grateful for any suggestions regarding
OT alternatives.

If you have any more questions, you can call me at 1 650 691 2250.

I'd like to hear more about the UVA system, too.

-- Rich Fuchs [log in to unmask] (Research Libraries Group)