“In JSF, the state of the each component is stored in between the requests” – This means that the data is cached(!?) at the application servers thus the servers will now have larger memory requirements(!?).
But I guess we can also make the state requestScoped, so this will not cache the data at application servers during the two requests, right? Am I getting it correctly ?
[My application contains blog posts of users which are large text blobs which would be very costly to cache. But yes there is some data that requires caching! ]
The sentence you mention here is about JSF state saving, which saves the state of the component tree. By default, this is done on the server… but you could also do it on the client, to save some mememory. See this: http://wiki.glassfish.java.net/Wiki.jsp?page=JavaServerFacesRI#section-JavaServerFacesRI-WhatAreTheDifferencesBetweenServerAndClientSideStateSavingAndWhatAreTheBenefitsDrawbacksOfEach
Backing beans scope is something else. You can make backing beans session / request / view scoped (or even use another or custom scope). This will have an impact on server memory usage. However, if you do it correctly and don’t have big data requirements, you could manage that. For example, you could store as little as possible and reload your data regularly (from db, …).
All this doesn’t mean your data is cached (at least, not the data coming from persistent storage). This is up to you.
So, memory requirements are something to keep in eye. The basic setup for a full-blown Java EE server is typically more than for a php application on an apache server… But, done correctly, making it faster / better / more scalable (cpu / mem) could be simpler to achieve.