HTTP Interface to memcache

2009-04-02 Thread Vijay Raghunathan
Has there been any discussion on exposing a REST style interface to memcached? It seems like a really natural way to access the cache since it already supports similar notion (GET/POST/PUT/DELETE would map very well to get/set/add/delete). Additionally clients are already written in every languag

Re: HTTP Interface to memcache

2009-04-03 Thread Ray Krueger
> Has there been any discussion on exposing a REST style interface to > memcached? At one point I wrote a web app that had a REST interface to run out in front of memcached. The idea being that the REST front-end could be load balanced and configured such that clients only need to know one addres

Re: HTTP Interface to memcache

2009-04-03 Thread John Allspaw
I remember being the "crazy old coot" and asking this same question a bit ago: http://lists.danga.com/pipermail/memcached/2004-November/000838.html BrianA managed to hack something out that is along the same lines, but I haven't used it: http://tangent.org/608/mod_memcached.html -j On Thu, Apr

Re: HTTP Interface to memcache

2009-04-03 Thread Josh Snyder
This is a cool idea. However, one downside is that it will increase the payload required to get/set a key, sometimes significantly. Memcached is usually used for performance and scaling, and one of the issues one encounters down that road is network load. We are extremely latency sensitive, and w

Re: HTTP Interface to memcache

2009-04-03 Thread Vijay Raghunathan
Hmm ... according to wikipedia(1), POST means Create (so I guess add in memcache terms) and PUT means Update or Create (so set). It is weird that POST knows the resource name. The way that CouchDB handles this is pretty clean(2) (imo). POST means create a document with a server generated key (n

Re: HTTP Interface to memcache

2009-04-04 Thread bokmann
You could literally write this in about 16 lines of code using Sinatra, a ruby dsl for creating web apps. I'm a bit busy for the next few weeks, but when I find some downtime, I'd take that on. I think it would be better done as something separate from memcache ayway - I don't need it trying to