Meeting Agenda Meeting Agenda For most of its life the Macintosh has lacked both capabilities. As a result, it has been difficult to automate repetitive interactions or to solve large problems on the Macintosh by using a multiplicity of applications as tools.
AppleScript is changing that. We'll take a look at AppleScript, and how it was used at the National Research Council to solve a database synchronization problem by coordinating "Custom-Off-The-Shelf-Software" (COTSS).
The University of Waterloo
200 University Avenue
Waterloo, Ontario
The Institute for Computer Research
Presents a Colloquium on
Automatic Speech Recognition Technology in Telephone Networks
by: Dr. Matthew Lennig
of: Senior Manager, Speech Technology and Applications
Bell-Northern Research
Date: Thursday, December 7, 1995
Time: 2:30 pm.
Place: William G. Davis Computer Research Centre, Room 1302
Abstract:
Speech recognition technology has seen tremendous growth in the
past decade and will continue to expand its success in wide rang-
ing application areas including telephone networks. In this
presentation, history of speech recognition technology with ap-
plications to telephone networks will be reviewed, followed by
outlining a few most successful applications Bell-Northern
Research has achieved in the past decade which have received
world-wide recognition and have excelled in world-wide competi-
tion in the telecommunication industry. These successful appli-
cations include the recent voice-connect project and the flexible
vocabulary speech recognition technology deployed for automated
telephone directory assistance and for real-time automated
stock-quote information retrieval. Demonstrations of some of
these applications will be given in video in this presentation.
Everyone is welcome. Refreshments served.
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
DATABASE SEMINAR
-Friday, December 8, 1995
Gopi K. Attaluri, graduate student and PhD candidate,
Dept. Comp. Sci., Univ. Waterloo will speak on ``Two-
Phase Commit Protocol in Transaction Processing
Systems''.
TIME: 1:30-2:30 p.m.
ROOM: DC 1304 *NOTE ROOM CHANGE*
ABSTRACT
Two-phase commit (2PC) protocol is often used by a
transaction manager to atomically terminate a
transaction accessing multiple resource managers (or
database systems). The transaction manager communicates
with all the resource managers and coordinates, while
the resource managers serve as participants. The number
of forced log records and the messages exchanged are
important performance factors of a commit protocol.
Several optimizations to 2PC have been suggested.
Transactions abort relatively rarely, therefore the
commit operation is improved at the cost of an
expensive abort. The presumed-abort (PA) minimizes the
forced log entries and presumed-commit optimization
reduces the messages. PA has been incorporated in
multiple database products and the X-Open distributed
transaction processing standard.
In this talk, I argue that PA is not scalable to a
hierarchy of resource and transaction managers, wherein
each node coordinates the 2PC protocol over its
children and in turn participates in the parent's
protocol. Here, PA can be employed only between the
root and its children. An intermediate node employing
PA could not support the prepared state required by its
parent. I propose a scalable 2PC optimization scheme
that matches PA's performance in transaction commit. A
PA-coordinator can easily switch to this scheme. Its
parent and children require no modifications, and need
not even be aware of the switch. A survey of the recent
literature indicates that the scheme may be novel.
Through this talk, I hope to discuss its soundness and
hear of any related research.