2003-11-02
by chris, 7719 days ago
"This case study discusses a project to build a dynamic web site and accompanying backoffice web application to maintain it built using non-conventional tools and techniques: Common Lisp and [create Object Prevalence]."an easy and interesting read (although the domain object model diagram that claims to be UML but is not, is a bit annoying :) especially interesting due to the object prevalence part. while by far not as dogmatic as other object prevalence writings, the belief that using object prevalence excludes the use of an RDBMS still surfaces sometimes. but in reality, combining object prevalence with an RDBMS is a viable and useful options.
prevalence systems require two persistent stores: a snapshot store and a transaction log. and why not store at least the snapshot to an RDBMS? with the help of an O/R-mapper this can be easily accomplished and storing data in an RDBMS is still a good choice for keeping data accessible and available. a separate system could also apply the transactions to the RBDMS-store and so the database would keep quite up2date without impacting the performance of the frontend systems. so combining object prevalence w/ an RBDMS is a perfectly fine thing.
one problem inherent to object prevalence (as it is proposed/used currently) is not adressed in this case study, as the system described is a single-machine deployment. but once you have multiple machines, how do you keep the in-memory data synchronized among the clustered machines? how to handle concurrent modifications? various possible approaches come to mind, however, if you are thinking about using a prevalence system in a clustered situation, those are questions that need to be answered.
search
29 active users
- unknown (29)
backlinks (more)
none, yet
recent stores (more)
recent comments
echo earlZstrainYat|tr ZY @.
