 News |
FAQ
|
| Search
|
Home
|
| Getting Started | Documentation | Glish | Learn More | Programming | Contact Us |
|
| Version 1.9 Build 1367 |
|
First, the Panel would like to thank the AIPS++ project staff for preparing a thorough and in depth review of the project status. We appreciated the clarity and candor of the presentations, and believe that we have been able to form a reasonably clear picture of the accomplishments, technical challenges, and programmatic difficulties of carrying out this ambitious project. It is clear that preparation of the review required a major effort on the part of the staff, but we hope that this provided an opportunity for everyone -- not just the Panel -- to step back from the day-to-day activities and assess the overall program.
Our most general observation is that leadership at the Project level is largely absent, and that technical leadership is informal and ad hoc. Despite this, the technical direction of the AIPS++ project has been forward-looking, sometimes aggressively so, unafraid of complex computing technologies, and abreast of or even ahead of the state of the art. The technical leadership has spent a great deal of its energy working out capabilities that might be needed by anticipated applications whose features and priorities are largely unspecified. In the absence of application-area guidance, the technical leadership has worked hard to define and implement infrastructure capabilities sufficient to support all of the envisioned or imagined application needs as inferred from the few requirements documents currently available and as synthesized from knowledge and experience in astronomy and computing. The result has been to make slow but steady progress on an incredibly broad front.
Working ahead of the state of the art is acknowledged to pose the danger of having to discard developed code and retrofit applications as industry standards appear and are adopted. This is seen as a necessary cost of doing business at this level of sophistication. Another potential danger is that the development of computing infrastructure can take on a life of its own. The issue of how far the computing infrastructure should go is an important strategic decision that has evolved to its present state by many defaults and the strong design work of several individuals and not by virtue of a strategic plan for this project.
The Project technical leadership claims to have a clear idea of the difference between infrastructure needs and infrastructure dreams, but the presentations and other discussions provide the Panel with no confidence that this vision is shared by all. The Panel's view is that there is in fact no basis for the claim that the real infrastructure needs are well understood by anyone because of the lack of guidance concerning the nature, ordering, and priority of astronomical applications to be developed in the AIPS++ framework.
The AIPS++ project has not been, and is not, managed in any conventional sense of the word, despite the existence of a project manager! The project management function to date has involved little more than monitoring what has happened, as the dictates of different players held sway. The project is essentially anarchic in character and out of control, driven in the main by egos, private agendas, and personal chemistry. There are no formal decision making processes, most decisions being justified on a post-hoc basis. There are no meaningful schedules, milestones, dependencies, or resource allocations,making it next to impossible to predict outcomes with any confidence whatever. There is no evidence of even the most elementary of project management methods and tools being utilized to manage the project. This is the case despite the expenditure to date of approximately 40 to 45 person-years of development effort at an estimated cost of some US$5 Million. It is not at all clear to the Panel that this quite considerable effort and expenditure has been as productively utilized within the AIPS++ development as might have been the case if the project had been well managed, rather than being allowed to ``follow its nose.''
In some sense it is unfortunate that the focus for AIPS++ project management has been NRAO, which does not have a culture of project management for software projects -- for hardware based projects external project managers are specifically retained. It should be recognized that much of the success of radio astronomy experimentation now depends directly and critically on software and, consequently, that the development of ``software instruments'' needs to be as carefully project managed as for hardware.
There is a remarkable lack of definition, in terms of functionality, of what actually is being built. Features are added or not, as the case may be, at the whim of the technical developers. The only requirements-like material that exists in documented form is little more than an all-encompassing wish-list. A partial architectural design document only appeared once this review was seen to be inevitable.
There is a sense in which the AIPS++ development is regarded as just another research project, which it certainly is not. It must be regarded as a serious attempt to build, in software, an important research instrument. Few people will happily accept the folly of conducting important basic research on what is essentially research infrastructure!