On Mon, Apr 2, 2012 at 9:39 AM, Stewart Smith <stew...@flamingspork.com> wrote: > On Fri, 30 Mar 2012 22:55:30 +0300, Henrik Ingo <henrik.i...@avoinelama.fi> > wrote: >> You could add some kind of authentication or security to HandlerSocket >> though, but the real reason this couldn't be a GSoC project is that >> HandlerSocket is not actually in Drizzle yet. So the only server-side >> GSoC project we would be interested in is to port Memcache API (rather >> than HandlerSocket) to Drizzle. > > Honestly, I think a full REST API is going to be a whole lot better than > ever having handlersocket itself. Why? It'll get used a lot more.
Stewart, REST API and HandlerSocket/MemcacheAPI are completely different things. REST API is "convenience" api. Easy access over http and key-value (or sql, even!) semantics. HandlerSocket and Memcache API provide "RAW" access, bypassing SQL parser, with the key objective of achieving better performance. They typically achieve 2x (for writes) to 7x (for reads) more, so it is well worth having them. Marcus' GSoC is a natural progression of this idea: Move the SQL parser (or part of it) to client side is good scaleout architecture. You get to keep SQL (and legacy apps) yet get better performance out of the server. HTTP is poor choice performance wise and of course in json_server we use Execute API so at least for now it "adds back" all the SQL overhead anyway. henrik -- henrik.i...@avoinelama.fi +358-40-8211286 skype: henrik.ingo irc: hingo www.openlife.cc My LinkedIn profile: http://www.linkedin.com/profile/view?id=9522559 _______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : drizzle-discuss@lists.launchpad.net Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
