Re: Gopher, trapped in amber

Robert S. Thau (rst@ai.mit.edu)
Tue, 12 Jan 1999 22:06:13 -0500 (EST)


Kragen Sitaker writes:

> > If you're interested in looking at some of the other fish in the pond
> > at the time the web piranhas took over, a few interesting things to
> > look at would be:
> >
> > *) WAIS, a Z39.50 variant.
>
> I think it's the other way around. The protocol and software were
> originally called WAIS; Z39.50 was the official (ANSI?) standard; then
> the sponsors of WAIS started asserting their trademark, so people
> started calling WAIS Z39.50.

Sorry --- Z39.50 came first. I'm very well positioned to know; WAIS
came out of text-database work done at Thinking Machines, well over
ten years ago at this point, back when I was working there, which I
had something of a role in.

Briefly: one of the first demos of the original CM-1 Connection
Machine was relevance feedback text database search; this demo was
held over and extended for the CM-2 introduction --- which is where I
got involved --- and was actually commercialized and sold to Dow Jones
as part of a dial-up service. The original WAIS work was led by
Brewster Kahle inside Thinking Machines as a project to figure out how
this sort of service could be delivered over the Internet (I *think*
it was no longer the Arpanet at this point). When they were looking
for a standard protocol to glom onto, Z39.50 was the closest thing
they found.

(But, as I understand it, Z39.50 is kind of like SNMP in that the
servers each export different data structures which the clients have
to know about --- WAIS servers all support one such common view, which
is why I described it as a Z39.50 variant. There are plenty of Z39.50
data sources out there that can't be queried with a WAIS client).

This work got spun out of Thinking Machines when it became clear to
just about everybody that almost nobody has a text database large
enough to need a Connection Machine, and the few people that do are
rarely adventurous technically. Brewster and friends sold out a few
years later (IIRC, AOL was on one of its inexplicable buying sprees),
and made quite a bit of money on it. Unfortunately, at this point, I
was *not* in the picture...

The CM-2 demo was more than ten years ago. Suddenly, I feel old.

> > *) Prospero, whose sponsor (Clifford Neuman) was trying very
> > hard to make it the One True Protocol for net information access.
> > See http://gost.isi.edu/info/prospero/
>
> Thanks for the ref. I wanna make a OTP too. :)
>
> I believe archie has been based on Prospero since around 1993 or 1992.

I know someone who did some work on Prospero, who thought this was a
really big deal, and much improved prospects for widespread adoption
of Prospero. I tried to be polite about expressing my skepticism, but
I probably didn't succeed.

What ever happened to Archie, anyway? Is it still running?

> > WAIS is the only one of these which ever saw in really widespread use,
> > though both the other two had a lot of political push behind them.
>
> So did OSI. :)

Ummm... Z39.50 *is* part of the OSI protocol suite, with BER and
everything, even if almost nobody is running it these days over OSI
transports.

rst