VISTA Enterprise Network - Successful Implementation, World Class Support

Thursday, June 25, 2009

Point 1: VISTA Requires the VISTA Software Lifecycle

Dear Reader,

Over the thirty-two years of VISTA development, many, many different software lifecycles have been tried with it, and all but one of them have failed, sometimes subtly, sometimes spectacularly.

Interestingly, for the last fifteen years no one has consistently tried to follow that one proven model, and during that time every VISTA adopter has struggled with VISTA. Those who have deviated the least from the model, like Indian Health Service, have enjoyed the most success, and those who have deviated the most, like the Department of Defense (DOD), have suffered the most. Veterans Affairs (VA) makes the best test case to prove this point, since they have been at their most productive with VISTA when they followed the model, and at their least when they didn't.

From studying VISTA’s rich and varied thirty-two-year history, I draw this radical proposition: we should take the plunge. We should end the fifteen-year drought by completely following VISTA's own software lifecycle model, the one that worked.

Getting to know the VISTA software-lifecycle model will take time, because it is sophisticated, complex, undocumented, and no example of it exists today, but I am confident that the more you get to know it, the more you will come to agree with me that the weird qualities of this model exactly support the weird qualities of VISTA in a way no borrowed or adapted model ever can.

Of these first eight points, this is the simplest, the most important, and the least likely for anyone to believe. I wish that weren't true, because this is the crucial missed point at which most VISTA adopters went off the tracks since 1994. If you can resist the urge to "improve" a model you do not yet understand, if you can compel yourself to study it patiently as though it were complex enough to deserve your attention, then you can buck the odds and reap the rewards that come with it.

Yours truly,
Rick

4 comments:

R. Kay said...

Should we consider looking at the last "clean version" release of each package, and then begin our discussion of the life cycle from that point?

Rick Marshall said...

Dear Rodney,

Good point.

The decision about whether to revert some packages back to their last clean version will probably have to be made package by package.

Some, like File Manager, can build off not only the most recent release, but off work done since then (in the case of Fileman, by George Timson for Medsphere) and not yet adopted by VA.

Others may well need to go back to an earlier version before they were architecturally damaged. I can think of no examples offhand, but I do remember that there are such cases.

Eventually, I agree we should investigate in this discussion how to launch each package properly.

In the meantime, I also agree we will need to discuss the general principles and problems involved with taking an overly centralized codebase and reorienting it to the VISTA lifecycle.

Yours truly,
Rick

R. Kay said...

As a start, I have expounded on a discussion you and I had a few years ago. I have uploaded a Word document to my Huddle workspace on LinkedIn and have sent you an invitation to take a look.

Rick Marshall said...

Dear Rodney,

My apologies for the delay in responding; 2009 has been the most disrupted year of my life so far, with six deaths in my family.

I'll go check out the Word document and start getting involved in the Linked In group you put together. Thanks for the pointer.

Yours truly,
Rick