Jump to content

Talk:Causal consistency

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Error and sources

[edit]

The current version of the page was created during "exam cramming" by the author and was just reformatted. The used source is not authoritative at all, contains a mistake and has a likely copyright violation (it seems more or less copied from Tanenbaum's "Distributed Systems", 1st ed., §6.2.3.); indeed, even Tanenbaum's book contains the same mistake (or at least, lack of clarity). As given, the definition does not imply that causal consistency implies PRAM consistency. So I'm removing this link: http://cs.gmu.edu/cne/modules/dsm/orange/causal_con.html

The explanation about why subsequent writes are causally related is an intuitive explanation that I created on my own; however, the new definition is correct. --Blaisorblade (talk) 16:03, 27 May 2009 (UTC)[reply]

I agree that the definition as given lacks clarity. In particular, it is not clear to me whether causal consistency implies that a read following a write w on the same thread must read the result of w. In other words, from the definition, I do not see that read-your-writes consistency needs to be given, if causal consistency is given. While the provided intuition why to subsequent writes on the same thread are causally related indicates that read-your-writes consistency needs to be satisfied, the definition itself does not contain this requirement. Does anybody know the details here, and can clarify things? —Preceding unsigned comment added by 82.130.103.118 (talk) 08:20, 9 December 2010 (UTC)[reply]