News for hackers. Stuff that matters.
Re:Local State Is Poison (2012) by im3w1l on Friday December 27, 2013 @10:04AM (#6970813)
I don't understand anything this article is saying. What is RDP? What are "live programming, open extension, and metacircular staged programming. "

What, and why would you want to upgrade during a loop?

>Those concepts are replaced by discovery – potentially in an infinite graph of stateful resources

What?

>To achieve large scale, robust, resilient, maintainable, extensible, eternal systems,

So many buzzwords...

Could someone give an example with before/after-pseudocode. I understand neither what the problem is nor the how the proposed solution is expected to solve it.

Re:Local State Is Poison (2012) by stcredzero on Friday December 27, 2013 @01:55PM (#6972025)
What, and why would you want to upgrade during a loop?

VisualWorks Smalltalk VMs had this ability, provided you were a bit careful about how you modified the system. The system would update atomically, and if you were doing things like adding instance variables that safely defaulted values, the system could update in the middle of transactions, then complete them with the new code base. There was the caveat the adding an instance variable to a class meant that every instance of the class basically had to be recopied with the new slot, which could potentially pause the VM for a significant span of time while it performed this operation.

There are embedded RTOS that have similar capabilities. Sometimes, you just want to get very close to zero downtime.

Re:Local State Is Poison (2012) by arethuza on Friday December 27, 2013 @10:14AM (#6970854)
To be fair, there is an "About RDP" link on the page:

http://awelonblue.wordpress.com/about/

Re:Local State Is Poison (2012) by mattmanser on Friday December 27, 2013 @11:23AM (#6971213)
I personally find it extremely telling that there isn't a single line of code in the buzzword filled page.

The entire tree concept he is talking about seems like a misplaced faith that you can achieve that kind of separation.

My gut says that in moderate to large sized programs so much state would end up under the root, just out of laziness, time constraints, over-complexity or lack of programmer skill, that you'd have the worst of all worlds.

I just caught the end of the global variables era and have worked on a few programs where the state of things can get modified anywhere.

It is not pretty.

Re:Local State Is Poison (2012) by jamii on Friday December 27, 2013 @12:18PM (#6971516)
While David Barbour does tend towards thinking about loud a great deal before coding (which is certainly sensible), he has also previously implemented these ideas in Sirea (https://github.com/dmbarbour/Sirea) and has just released a spec and interpreter for his Awelon language - http://awelonblue.wordpress.com/2013/12/25/awelon-project-pr...
Re:Local State Is Poison (2012) by kbenson on Friday December 27, 2013 @11:31AM (#6971268)
I've worked on similar code. Global state which is sometimes twiddled directly in the main loop, other times functions are called that take no arguments and return nothing, but they compute something and set global variables.

Not pretty is an understatement. It's ugly at best, and rage inducing most the time.

Re:Local State Is Poison (2012) by dllthomas on Friday December 27, 2013 @01:49PM (#6971997)
It helps when you can use the type system to restrain what can touch what. I've been doing this in some of my high performance C, where some global mutable state is unavoidable - passing around empty structs to indicate context. It doesn't guarantee I don't violate my rules, but it more likely I notice when I do.
Re:Local State Is Poison (2012) by kbenson on Friday December 27, 2013 @02:09PM (#6972089)
It's almost never as bad when you design it yourself. Then the logic and reasons are apparent. Coming into such a system from the outside can be tough though. An explanation of the rationale and pre and post conditions is very helpful, even if all it does is keep a future programmer from trying to circumvent the system and causing the problems it was implemented to avoid.
Re:Local State Is Poison (2012) by yalue on Friday December 27, 2013 @01:47PM (#6971986)
For me, the problem wasn't only this, but also that RDP is already a well-known acronym for "Remote Desktop Protocol".

One of my personal policies is to not introduce any new acronyms in the work I do, and especially don't redefine ones that are already familiar in the programming field. Just come up with a proper descriptive name, even if it's slightly longer. Let the acronym follow once the name is already known.

Re:Local State Is Poison (2012) by stcredzero on Friday December 27, 2013 @01:57PM (#6972035)
A friend of mine who worked at the EPA came from a culture where they liked to stomp on existing computer field acronyms. I even once worked on a big company system called IO.
Re:Local State Is Poison (2012) by yalue on Friday December 27, 2013 @02:12PM (#6972097)
I developed my distaste for pointless acronyms after working at a large company that had little regard for sensible terminology and would tack acronyms onto everything. Plenty of acronyms not only conflicted with established acronyms in the field, but also with others within the same company.

Seeing as how your friend worked for the EPA, maybe I shouldn't be surprised that this phenomenon often seems to surface within organizations whose name is typically given in the form of an acronym...

Re:Local State Is Poison (2012) by lhgaghl on Friday December 27, 2013 @01:45PM (#6971966)
> What, and why would you want to upgrade during a loop?

See Erlang for a trivial example.