Comments on High Level Framework for AITS Architecture
November 17, 1997
These are comments on the 18 Oct 97 draft
of the foil presentation High Level Framework for AITS Architecture
sent by Rich Ivanetich by email.
Top Level Comments
The presentation is needed and on the right track but will require more
work to be fully useful.
The presentation represents one (and not the only) very useful view
of the AITS architecture, namely a view that focuses on architectural properties
that the AITS architecture must have to achieve Rick's 4 Cs (complete,
consistent, correct, and current) plus Rich's E (evolvable).
This is a good level for the Architectural Review Team to ponder.
One of the questions asked of the review team was, "What role can the Review
Group play in terms of the long-term framework?" If this document
is related to that question, then its worth making these additional comments:
The High Level Framework is not a long term architecture -- it is
needed now. And all AITS programs need to begin to understand
this view of the architecture and contribute to a group understanding of
This is also not the only view of AITS that the Architecture Review Team
should contribute to. We should also be identifying technologies
that might cause discontinuities, identify process questions about how
to get componentware into wide spread use, and generally provide expertise
to improve AITS in the most useful high level ways we can.
I hope this critique is useful. I agree with most of your conclusions,
think both sense of framework are useful to AITS, and so like the spirit
of the presentation. I am just not sure the reasoning process to
the conclusion is as clear as it might be. Having thought about this,
at this point, I think the sense of framework that is most solid in this
presentation is the second one and maybe only a few changes would be needed
to make this clearer, if that is what you intend.
The word framework in High Level Framework is used in two senses
making it ambiguous and confusing.
Both senses are useful but they are different. Both are areas where
the Architectural Review Team can help.
as a meta architecture, that is the perspective of architectural properties
that all components must abide by. Most of the discussion on
ilities is apropos to this sense of the word. But describing the
framework as a "level above the architecture" is then confusing since it
is still architecture.
as an organizational, process, or presentational device to insure that
we look at the relevant aspects of the architecture. The identification
of the architectural capabilities and then their use in foils to show Trends
in Architectural Capabilities, Comparison of JTF and ALP, and Proposed
Enhancements are examples of the use of a framework to draw out different
information or consider different aspects.
The category scheme is very vague and high level to me.
But I do agree this is still somewhat useful to try to do in some cases
like Information Representation, to see trends like the move toward finer
granularity and flexible composition, and to compare roles played by objects
and agents and CORBA versus RMI (which do not show up if you just consider
Behavioral Control, Component Interaction, and Component Composition especially
overlap. So do Information Representation and Semantic Expression,
the former seemingly a subset of the latter.
The category scheme seems to be descriptive and not really constructive,
leading me to view it as more the latter kind of framework.
I'm not sure how easy it would be to map the wide range of technologies
on all the AITS slides into these categories, whether any two people would
agree, and whether the result would be useful.
I do see it as being very valuable to identify the various ilities and
analyze them. By their nature, ilities cut across systems and one
wants a way to insert them into systems that evolve and to control their
policies. I see us doing this for reliability, network management,
scalability, composability, ... mostly separately now built in to systems
in a brittle way, but we are beginning to see the pattern for inserting
these into composable systems. That is what my OMG
OMA NG paper and the upcoming January
workshop are leading toward.
on the foil Dimensions for Extending Current AITS Architectural Capabilities:
I agree on all the areas of extension but am not sure I agree that the
framework of dividing the world into Object Management Services,
Agents, Components, and System Control Services got us there or that the
categories are non-overlapping. So it might just be useful foilology,
not a meta architecture in the first sense.
on the foil Selected Technology-Enabled Enhancements, I have a similar
comment. I agree the enhancements are reasonable but am not sure
the high level framework really got us to this conculsion. (Yet another
meaning of framework seems to be creeping in here in this foil,
one that is sort of synonymous with DSSA, that is patterns for subarchitectures
of the whole.)
on the foil Proposed Enhancements, similar comment
Hope this helps,