.comment-link {margin-left:.6em;}

filling the void

Wednesday, October 03, 2007

Choice Overload

I found an interesting blog post over at Intel's research blog titled Parallel programming environments: less is more.
It talks about how offering users too many choices that individually have little or no benefit over the others might cause users to walk away instead of diving in to the sea och choice and working something out.

My comment to the article was as follows

I would say that this applies to more areas of programming than just parallel programming. I think that Sun made the correct decision when it delivered not only a programming language for java but an incredibly complete API to go with it.
If you compare to C/C++, which is similar in performance and use today (people will argue about that I guess, but there are many studies both confirming and denying what I just said, anyway...), you get a similar language with similar language constructs, but you have a horde or APIs to choose from for virtually anything you could possibly want to program. Everything from string handling to sockets to threads.
Much the same way with linux distributions and software in general.
Consolidate choices to make fewer and better products. (This only works to some exten[t] though, much like uniting countries in unions (us & eu, for instance))
This is a topic that often comes up in conversation for me. Not with regards to parallel programming, but rather in connection with the things I stated above, software in general and countries or organizations. I realize that someone may have a unique vision and decides to create a product based on it, and this is fine. The problem is when everybody does it, instead of picking one of the already existing products and helping to make it better.
I believe that choice is good, choice benefits the consumer if there is a surmountable amount of it. There is always a breaking point, where uniting too many things will reduce the entire effort to a rubble with too many creative minds wanting to go in different directions, but I firmly believe that, taking the world of linux distributions for an example, it would be possible to create something like 3-4 distributions with heaving intellectual and professional backing, rather than the around 350 virtually identical distributions we have today.
Much like Microsoft went overboard with (I think) 7 different versions of vista. Microsoft has succeeded so far for many reasons, but one of them have been by limiting the choice the users have. Getting everybody to rally behind one cause inevitably makes that cause better than having a single cause for every participant.

Diversity is good, but having too much of it hampers development.

(I realize that I've been using the Swedish spelling of parallell (with the double Ls at the end) all over the internet, including in the Intel blog comment. I'll try to keep the languages apart from now on.)

UPDATE: Apparently Intel is very selective about what comments it allows in its posts. I should have realized that I was not the first to comment on a 1-day old story that figured on slashdot, Intel just isn't showing any comments...

Labels: , , , ,


Post a Comment

<< Home