Allen Gilliland wrote: > > > Elias Torres wrote: >> >> Allen Gilliland wrote: >>> Looks fine to me, so +1, but one note. we should be clear about the >>> caching strategy. my expectation is that we would not cache these feeds >>> just like we don't cache search results pages. that would need to be >>> accounted for in the FeedServlet. >> >> It makes sense. It makes so much sense that I'm writing up another >> proposal for complete reverse proxy support (e.g. no page cache at all, >> optionally). >> >> We would like an option that disables page caching in the app server >> because we discovered a large amount of CPU being consumed by >> StringBuffer memory management instead of writing directly to the socket >> stream. > > Sure, I'm always interested in supporting more caching configs =) > > You *can* actually disable all of Roller's rendering caches very easily > in the current code. Each of Roller's caches supports an "enabled" flag > in the config, so if you put ... > > cache.weblogpage.enabled=false > cache.weblogfeed.enabled=false > > in your config file then you don't have caching in the app anymore. i > use this feature regularly in testing.
Right on. > > -- Allen > > >> >>> Also, we provide the option for the built-in search to be disable, and >>> that would also need to be handled in the FeedServlet so that search >>> feeds are not returned when search is disabled. >> >> Right. Good catch. >> >> -Elias >> >>> -- Allen >>> >>> >>> Elias Torres wrote: >>>> Hi guys (again), >>>> >>>> Here's another simple/short proposal that I would like comments on so I >>>> can get a sense from the community if I can proceed and when. >>>> >>>> http://cwiki.apache.org/confluence/display/ROLLER/Proposal_SearchFeeds >>>> >>>> -Elias >
