Talk:Shared-nothing architecture

Not just databases
I notice in the history that there has been some question about where this term is applicable. It originally referred just to databases, as shown in the Stonebreaker citation, but the SN term is no commonly used for PHP and Ruby on Rails web applications even when there is no database involved. The category probably needs to be broadened, but I'm not sure what to use. David 17:06, 26 April 2006 (UTC)

Rails?
I question that Rails is a shared nothing architecture. By default, it has file-based sessions, as well as a single process runner which doesn't get reset between requests. If you set a global variable in one request, it'll be there in the next (assuming you end up at the same backend). --Happygiraffe 07:45, 1 October 2006 (UTC)

Inappropriate examples
--Dnjclarke 19:16, 21 October 2006 (UTC) I agree with Happygiraffe and David. PHP and Rails are not appropriate examples of Shared Nothing architectures, and J2EE is not an example of the opposite. Programming languages are not architectures. They may lend themselves to implementing certain distributed architectures, but all the examples are flexible enough to implement any kind of architecture, be it shared memory, shared file, or shared nothing.
 * It's about a year later; I removed the (growing list) of popular du jour frameworks, API's and/or programming languages from "examples" of SN architecture. (Specificially removed this line: Three popular web development technologies, PHP, Django, and Ruby on Rails, all emphasize an SN approach, in contrast to technologies like J2EE that manage a lot of central state.  (One wouldn't say  that C or C++ "emphasizes shared-nothing, even though many (most?) SN arch implmentations are often actually implemented in C/C++?) Michael (talk|contrib) 21:59, 30 June 2007 (UTC)

Naming (a linguistics aside)
Alternative names: My issue is that, when I first ran across this page, my initial thought was "it's an architecture for resource sharing, but the resource that it's sharing is... nothing...?  That... doesn't... make sense, even if there are cases where 'nothing' is really used to denote something.".
 * "shared nothing architecture": the current name here, it gets quite a few ghits
 * "shared-nothing architecture": a decent number of websites use this
 * "nothing shared architecture": 1 ghit

I'm not suggesting that any of these are bad names in any way, just that it's a linguist's job to figure out if slightly modifying the wording might help prevent even initial confusion.

I think the first two options are best because they maximize the distance between "shared" and "architecture". For some reason, despite the word in between them, I was still incorrectly parsing it as "(shared architecture) ... (nothing)", as strange as that sounds. So maybe adding the hyphen would make it clearer that readers can't separate "shared" and "nothing", that it must be parsed as "((shared nothing) architecture)"? --Underpants 15:30, 27 February 2009 (UTC)
 * Linguistics is inappriopriate here - it's a technical neologism (no doubt that will now attract the idiots who deleted Web 3.0) and we should follow the lead of its orginal coinage, which is "shared nothing". We already do this, so what's the problem? Etymologically, the root does trace (if from anywhere) as being a counter-example to the alternative architectures of "shared memory" and "shared disk". It's not an ab initio coinage, in which case it might well have been "nothing shared" instead.
 * It's also worth noting that this is a database architecture and not the cryptographic use of "shared nothing" in relation to key exchange - although that's also universally described as "shared nothing". Andy Dingley (talk) 17:44, 27 February 2009 (UTC)
 * a shared-nothing architecture would be an architecture where the nodes share their nothing, which is nonsense. "shared nothing" makes sense in the domain it lives in as a term that constrasts with "shared memory" and "shared bus" and "shared pool" and other "shared X" architectures where the X is often protected against being a potential single point of failure with redundancy.
 * I completely disagree with Andy. The hyphen in this case is syntactic, not semantic. It's there to distinguish modifier modifier object from modified-modifier object at the velocity that the eye makes parsing decisions. In formally punctuated English, one writes "shared memory solves many problems", but also "shared-memory architectures are a PITA".  Whether "nothing" is shareable or not has nothing to do with this as the hyphen is not semantic.
 * (My use of "formally punctuated English" was a bit of a in-joke, as the rule in English is to leave out the hyphen when the first modifier ends in -ly which already makes the word 'very' conspicuously adverbial, and therefore it can only function as a modifier of the adjectival "punctuated" leaving no parsing ambiguity for the eye to reconcile.) &mdash; MaxEnt 06:42, 28 January 2016 (UTC)

Examples
This article has no examples section. I would like to nominate starting such with Bitcoin. — Preceding unsigned comment added by 75.87.129.242 (talk) 22:25, 21 March 2012 (UTC)
 * on the other hand, bitcoin is something of a shared bus architecture -- it seems that whenever anyone says "shared nothing" they are overlooking whatever is actually shared, and if they stepped back further they'd find it. Shared power outlet, shared planet, blah blah blah 75.87.129.242 (talk) 05:27, 15 April 2012 (UTC)