More Continuation stuff
Monday, September 8, 2003
Yesterday, I noted that Chris Double is making progress developing a SISC Scheme continuation based web framework (as a learning exercise) and said that it'd be interesting to see what he had to do to separate out the "non-serializable stuff" in the continuations that he persists. Well, today Chris posted the detail on his blog as well as a summary of some of the issues that he encountered when moving from storing the continuations in hashtables to storing them on disk. (Thanks Chris!)
Avi Bryant followed this up with a list of some of the possibilities that you have for snapshotting and dealing with state. Again, a very informative posting! I have found the material that these two have been posting very valuable in understanding the issues and approaches for dealing with session state on continuation based web servers. The academic papers I've been reading address some of these issues from a high level, but Chris & Avi have been more specific about detailing approaches and complications. Very refreshing.
On a related note, I've also just come across a paper titled Scheme in the Real World: A Case Study that describes how some UK consultants used SISC Scheme as part of a three-tier system they developed for a commercial web application. In it, they discuss some of the continuation/session state issues and how they handled them with SISC Scheme as well. The tiers and the technologies they used for each tier were:
- presentation: Cocoon, Castor and XSLT
- business logic: SISC, JavaBeans, Object-Relational Bridge
- database: SQL Server
