So, Ive seen several people on the left speculating the failure of an INCREDIBLY COMPLEX PIECE OF CODE, STARTED LATE, WITH A FIXED DEADLINE, BUREAUCRATIC OVERSIGHT FROM HELL, AND EVER-CHANGING SPECIFICATIONS was some sort of Fiendish ReThuglican Plot To Sabotage Obamacare, as opposed to being about as unexpected and unpredictable as Celebrity couple breaks up. Ah, but now, Sarah Palin, showing the same deep knowledge of computer programming and project management that she has of foreign policy or completing your term as Governor, tells us the TRUTH: The failures were deliberate, but it was a DEMOCRAT plot! Ah-ha! They cunningly sabotaged their legislative showpiece, just to... something! Both of these non-theories (if you know what a theory is, you know that conspiracy theory is pretty much an oxymoron, believed by ox-headed morons) fail due to people needing to believe that hundreds of programmers -- probably adequately skilled and reasonably moral, no more or less than anyone else -- would keep silent about being deliberately ordered to fail at a task that will show up on their resumes from now until doomsday (which, depending on who you believe, is either the next Republican President or the next Democratic one). No, this project failed for the reasons most projects fail. Lots and lots of people probably tried, if not their *best*, at least as hard as most people try when theyre drawing a paycheck, and probably everyone involved believed they were doing the right thing, or at least, not doing a bad thing, and hundreds of people x too little time x too many mandates x insufficient testing == What Weve Got, and what anyone should have expected. If there is moral failure (as opposed to just good ol everyday failure) anywhere along the chain of responsibility, it is with those who failed to say This cant be done with the time and resources -- mostly the time -- allotted. (Throwing money at a complex problem isnt a solution. Moar c0derz! is rarely the answer -- it usually makes things worse. Nine women cannot produce a baby in one month, despite what project planners seem to believe.) According to this (npr.org/blogs/health/2013/10/22/239197047/how-politics-set-the-stage-for-the-obamacare-website-meltdown?sc=ipad&f=1001), real work did not begin until after the 2012 elections. Thats less than a year. The feature set AND the deadline were immutable (but the program had to be designed so as to be easily changed as regulations change). Public or private, more money or less money, it would not have mattered. The only way to have gotten something working in that short a span would be... never mind. There really is no way. You cant strip out features, because theres too much the program has to do to be at all useful. It could not be done, and someone should have said that, and if the government wasnt going to listen, said it publicly. I work with software that needs to address/be aware of a far smaller set of regulations, that is targeted at a highly knowledgeable audience, and where hardware and operating environment are all strictly controlled, and I know how complex it is. I dont need to look at the internals of the code to guess what a nightmare it must be. Even the bits and pieces Ive seen speak of something rushed to the public with virtually no real user testing on live data. A simple example: Pop-up menus with hundreds of items (Yes, really -- Ive seen the video.), not sorted. Wanna bet the coder had a test set of maybe 5 items, with unit testing to test 0, 1, and n items? For debugging, thats fine: If it works correctly with 0, 1, or n, the code probably works[1]. This doesnt mean the code is *usable* if the actual data is hundreds of items, even if it doesnt actually crash. You dont use a popup for lists that long. This would become instantly apparent if there had been time for a sane testing cycle with actual users, especially those who didnt know what to do -- people who know the right thing to do with a program will not catch the hundreds of wrong things a normal user will find instantly. [1]Probably. If you dont know how much data you have, and you test only with small data sets, you expose yourself to possible errors when you pass assorted magic numbers that are usually powers of two. (I have also found problems if you send data as a big chunk of text, because various APIs and interfaces might not deal with well with text over, say, 32K for some, 64K for others, etc. This may never show up with test data used by programmers.)
Posted on: Tue, 22 Oct 2013 16:03:00 +0000
Trending Topics
Recently Viewed Topics
© 2015