Re:Ask HN: Is Accidental Complexity Growing Faster Than Esse...
by arethuza on Monday March 27, 2017 @04:12AM
(#13965948)
Your simple example of associating a Customer/Client to an Address is a great one - it could be that a really simply approach is required, or you may actually need the powerful features around such things that some high end ERPs support.
In a previous job I spent a lot of time trying to get a shared understanding of what "customer" actually meant (this was initiated by realizing that the Customer Loss Ratios that execs wanted were really nonsense). This eventually turned out to be a non-trivial problem.
Re:Ask HN: Is Accidental Complexity Growing Faster Than Esse...
by beaconstudios on Monday March 27, 2017 @04:12AM
(#13965949)
that sounds more like a symptom of forgetting to include refactoring in your agile loop, than an inherent issue with small iterations.
Re:Ask HN: Is Accidental Complexity Growing Faster Than Esse...
by d--b on Monday March 27, 2017 @04:31AM
(#13966009)
Yes, in this tiny example, it could be avoided by adding a simple refactor in your second iteration.
In general though, the customer may ask you some iteration that he doesn't realize imply a complete rewrite of the code. Like:
"now I want the same thing, but I want it available 24/7 with no downtime".
Or "Now we have a great text editor, we want to add live collaboration".
"You know what would be great: that we could see what changed between 2 versions of the spreadsheet".
Re:Ask HN: Is Accidental Complexity Growing Faster Than Esse...
by beaconstudios on Monday March 27, 2017 @04:38AM
(#13966034)
true, and it's hard to work around unreasonable customer requirements. In the ideal circumstance you would have that conversation and explain that having 100% uptime guaranteed is basically impossible and aiming for e.g. 6-sigma uptime requires high ongoing costs - but I'm sure we've all experienced that "I can have it both ways" attitude.
As an example:
After a short development cycle, you created a database schema where each Client class has a single address, and because it was done quickly, the address fields are in the same Client table.
In a second development cycle, your customer tells you that their clients may have one or more address and one of them is the primary. The quickest path to delivery is to keep the address in the Client table as the primary address, and add a secondary address table.
While short development cycles help in not screwing up large requirements, it also impairs our ability to see the larger picture, and cause us to take decisions that we need to undo afterwards.