bt-spec message
[Date Prev]
| [Thread Prev]
| [Thread Next]
| [Date Next]
--
[Date Index]
| [Thread Index]
| [Elist Home]
Subject: RE: [bt-spec] BTP Issue 82 : ENROL related to application message -solution in 0.9.0.3
- From: Peter Furniss <peter.furniss@choreology.com>
- To: bt-spec@lists.oasis-open.org
- Date: Thu, 17 Jan 2002 00:53:06 +0000
My
contribution to the unclosed discussion on issue 82. (I've reverted the issues
status to "solution proposed" - I changed the issue colours again earlier this
week to make the votes show up better, so 82 is now maroon.)
There
is a link to the architectural question that Sanjay raised and I've put into
issue 65 - see my message on that. The linkage is that if you
allow propagation of a CONTEXT/cohesion, then the identification of
which bits of application work are associated with of the Composer's Inferiors
requires the relationship between ENROL and an application message arising from
the application work. It's less significant in the case of propagated
CONTEXT/atoms, because then the association between application work and
enrolled inferior can be done by which application message the CONTEXT related
to.
The
case is most readily explained by a scenario. I'll base it on the Choreology
demo at first.
The
application design is to ask each of several market makers (services) to provide
a binding quote on two items - a stock and an option. The quotes are "binding"
in that each is associated to a prepared Inferior (Participant). The
client end will compare the quotes and choose the best stock and the best option
- these need not be from the same market maker.
One
way of implementing this is for the client application to create a cohesion, and
pass that CONTEXT/cohesion with the "getquote" requests to each of the market
makers - with two separate getquotes to each. For each quote, the market
maker create a Participant, and enrols it in the cohesion, and returns the
particular quote information to the client.
It would obviously not work if the market maker just
sent an ENROL to the Composer and the quote to the client but did not indicate
back to the client+composer which enrol belonged to which quote. The
client, as terminator to the composer, can distinguish the different inferiors
(we have more than one means for this - "inferior handle" qualifier being one of
them), but needs to know which is which.
At abstract level, what has to be communicated from the
market maker to the client is that ENROL(23), say, is for our option quote, and
ENROL(25) is of the stock quote. The most obvious way to do that is to say that
ENROL can be related to an application message. That's all that the proposed
resolution of 82 was really saying : that it is a meaningful statement to say
that. The particular text in the related groups section is concerned with making
sure the ENROL goes near enough to the application to make this possible, given
that the Composer may not actually be colocated with the client application
(whereas the CONTEXT_REPLY's target is deemed to be sufficiently colocated that
the relation can be perceived).
----
The particular example can in fact be worked
differently (by making "local" atomic inferiors of the cohesion at the client,
and putting different CONTEXT/atoms on the getquote(option) and
getquote(stock). However, an obvious extension of the example won't work
that way, and needs the cohesion propagation
above.
If, instead of the client deciding to get both quotes
from every market maker, the decision of which quotes to submit is make by the
market makers, who are just informed that quotes are desired, then some market
makers may create one Participant, some two, and, if they perhaps wnat to make a
special offer, three. Or, given that this quotes are time-limited, they
might wish to offer, at the same time, a very competitive price with a short
timelimit, and a less good one, but with a longer timelimit. All of these
require some way of saying "the application work implied by this message (in our
case, a quote) has been put under the control of this BTP
Inferior"
---
there are some other yet more intriguing things you can
do down this path, but I think this gives the
essence
Peter
------------------------------------------
Peter
Furniss
Technical Director, Choreology Ltd
web: http://www.choreology.com
email:
peter.furniss@choreology.com
phone: +44 20 7670 1679
direct: +44 20
7670 1783
mobile: 07951 536168
13 Austin Friars, London EC2N 2JX
[Date Prev]
| [Thread Prev]
| [Thread Next]
| [Date Next]
--
[Date Index]
| [Thread Index]
| [Elist Home]
Powered by eList eXpress LLC