Message-Id: <199602090104.TAA04952@library.wustl.edu> Date: Thu, 8 Feb 1996 18:56:04 PST From: "Donald J. Reilly" <mailto:reilly@PSYCLONE.ENDINFOSYS.COM> Subject: Re: Loose Ends To: Multiple recipients of list WEBCAT-L <mailto:WEBCAT-L@WUVMD.WUSTL.EDU>
Chris,
I think that your confusion comes from a semantic issue. It is called
connectionless
because there is no connection between related interactions. Z39.50, for
example, describes a
dialog between client and server. The client says "Connect me to this database"
Server responds
"You're connected" Client says "issue this search on my behalf" Server
responds "Done, here's
the first 100 records" Client says "OK give me the next 100". There is a
connection, and more
important for this discussion, a context for each of the following transactions.
HTTP is
connectionless because it talks to everyone for the first time, every time.
There is nothing in
HTTP that allows the client to say, "Give me the next 100 records", because that
implies a
context to the request (a search, a result set, a pointer into the result set)
that simply doesn't exist.
So, one of the problems because of this is that in Z39.50 the server can say "I
have 10 users that
I'm talking to now". HTTP can only say "I've responded to 100 requests in the
last minute", but
does that mean 100 users or 10 users or 1 really fast user? HTTP doesn't know
because there is
no connection between the requests to link them together to be a dialog that can
then be
attributed to a number of users.
Thanks,
Don Reilly
---------------Original Message---------------
> The first posting to WebCat-L raised the issue of how vendors would
> license their web gateway products. At least two methods came
> forward: 1) pricing according to the number of concurrent "sessions"
> allowed, and 2) pricing according to the number of volumes in the
> client library's collection (analogous to software companies charging
> on the basis of the size of a central processor). A number of people
> pointed out that the problem with the first approach is that the inherent
> "statelessness" or "connectionlessness" of the HTTP protocol means
> there is no "session".
I don't think calling HTTP "connectionless" is quite right. You can
count/limit the number of simultaneous HTTP connections to a server.
There is a "session" of sorts: the client opens a connection
and asks for a url; the server sends it; the connection is closed.
You can think of it as very quick, anonymous file transfer.
When a browser asks for a CGI URL (e.g. the connection to
our z39.50 catalog service), it just sits and waits for the server
to send back "stuff". So in that case the connection lasts a lot
longer (a few seconds).
If you have a product that it licensed by simultaneous connections, it is
obviously more efficent to have the requests serviced as fast
as possible. The HTTP sessions just get the job done
and get out, unlike telnet or z39.50 sessions that are
controlled directly by a browsing human who sits and ponders
every screen (or wanders away without signing off).
--
Chris Howard mailto:choward@iastate.edu (515) 294-6521
Iowa State University Library -- Automated Systems Division
----------End of Original Message----------
*****************************************
Donald J. Reilly
Director of Sales
Endeavor Information Systems, Inc.
(708) 292-2292 x2604 Fax: (708) 292-2296
mailto:Reilly@endinfosys.com
*****************************************