On Wed, May 20, 2015, at 09:51 AM, Paul Hoadley wrote: > Hi Ramsey, > > On 20 May 2015, at 1:53 am, Ramsey Gurley <rgur...@smarthealth.com> > wrote: > > > On May 19, 2015, at 4:03 AM, Paul Hoadley <pa...@logicsquad.net> wrote: > > > >> I’ve had a look at Ramsey’s ERPersistentSessionStorage.framework, and my > >> intention was to create something similar backed by memcached. I got it > >> working to the proof-of-concept level on a trivial app, but using native > >> Java serialisation seems brittle. (By which I mean it falls over pretty > >> fast when you throw a serialisation-based approach into a substantial > >> application.) > > > > I’m curious about the "falls over pretty fast" part. What sort of problems > > do you have? > > Serialization failures. (To be clear, I’m talking about testing with my > own memcached-based approach, but it’s really only some very minor > changes to your work to use memcached instead of the database as the > back-end.) Initially it was my own classes, which I could fix, but then > I started hitting weirdness like this: > > > Caused by: java.io.NotSerializableException: java.lang.reflect.Field > > > It’s not obvious to me where that’s coming from, or how to fix it. I had > also formed the impression (perhaps erroneously!) that you had found > serialization issues deeper inside the WebObjects frameworks that were > going to be difficult or impossible to fix. > > > I had a few unresolved bugs that could have probably been fixed by > > rewriting EOEditingContext’s serialization routine, but I never bothered. > > That sounds a lot more promising. I was under the impression you had > abandoned the idea because the problem was unsolvable. > > > It became obvious to me that a) nobody was interested in fixing > > serialization in their legacy classes, > > I’m interested! > > > and b) everyone was too proud of their one-off direct action workarounds to > > this particular problem to look at a new solution that works for everything. > > I am really interested in being able to decouple session persistence from > the application server to allow for automated horizontal scaling of > component action-based apps. > > > Basically I wrote that framework to prove to myself that a component > > action/D2W app could be used as a “front end” app at scale and do a better > > job of it to boot. I succeeded in convincing myself that it was possible > > and left it there :-) > > > What would it take to get you back into this? Let’s start with a show of > hands: is anyone else interested in developing a robust solution to this > problem? > > > -- > Paul Hoadley > http://logicsquad.net/
I'm most certainly interested :) As you know, I dabbled with the framework for a project but came across some serialization problems deep in the frameworks. I am really interested in the approach you mention above, Ramsey, with the EC serialization routine. My intention was to use ElastiCache. -- Matt http://logicsquad.net _______________________________________________ Do not post admin requests to the list. They will be ignored. Webobjects-dev mailing list (Webobjects-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
