April 28th, 2015
(written by lawrence krubner, however indented passages are often quotes). You can contact lawrence at: firstname.lastname@example.org
This doesn’t rise to the level of “Zed Shaw rant” style of meltdown, but Jim Starkey has decided the play the role of “famous old codger who did great work once but now seems completely out of touch“:
Any discussion of the CAP “theorem” revolves around the A – exactly what does availability mean?
The narrow sense is that availability means that all surviving nodes in all partitions continue to process work. This reduces the CAP idea to nothing more than “there is no consistency without communication.” Well, duh. This is of interest only to third rate academics a few papers short of a tenure package.
Joe Emison writes the perfect reply:
The point of the CAP theorem–according to its author, Eric Brewer (I suppose a “third-rate academic” to Jim Starkey), was to point out that (a) for many use cases, availability is more important than consistency, (b) ACID databases sacrifice availability for consistency, and © there is a case to sacrifice immediate consistency for availability. And it took something like five years before there were enough other use cases in the wild (e.g., Amazon) to make these observations useful to serious practitioners.
Jim’s arguments are essentially, “I believe that I have a good way of sacrificing consistency for availability in a way that really minimizes the impact on consistency.” Jim, in case you hadn’t been paying attention, that’s essentially what every NoSQL database engineer has been trying to do for at least the past 10 years. You have just picked the most obnoxious way of saying it.
If hundreds of smart people have studied a problem for 20 years and still failed to come up with a solution, then it is probably inaccurate to say “This is of interest only to third rate academics a few papers short of a tenure package.”Source