Lighttpd as HTTP access layer for Jackrabbit

[tunggad <tu...@yahoo.com>]

Background for the idea:

- WebDAV Servlet and Tomcat as HTTP access layer is too heavy to serve a
large community.

What do you think about that idea ?

http://www.nabble.com/file/p20704613/lighttpd_jackrabbit.png 

-- 
View this message in context: http://www.nabble.com/Lighttpd-as-HTTP-access-layer-for-Jackrabbit-tp20704613p20704613.html
Sent from the Jackrabbit - Users mailing list archive at Nabble.com.

Re: Lighttpd as HTTP access layer for Jackrabbit

[Alexander Klimetschek <ak...@day.com>]

On Wed, Nov 26, 2008 at 5:45 PM, tunggad <tu...@yahoo.com> wrote:
> Background for the idea:
>
> - WebDAV Servlet and Tomcat as HTTP access layer is too heavy to serve a
> large community.
>
> What do you think about that idea ?
>
> http://www.nabble.com/file/p20704613/lighttpd_jackrabbit.png

With access only to the data store directory, the lighttpd servers
won't be able to look up the paths from the repository, as this
information is contained in the persistence manager's data. The
datastore is identified by ids (afaik these are the md5 hashes of the
content).

You can use a jackrabbit cluster node (w/ Tomcat) in read-only mode to
provide access to the repository, using Webdav. To improve
performance, you can put a caching proxy in front of it. You simply
need a mechanism to invalidate the items in cache whenever the
according contents of your repository change, but for the caching you
can use Apache with mod_proxy and caching enabled. This way you can
cache everything from your application.

A good and simple rule to separate dynamic content from cachable is to
cache only pure path URLs (eg. /mysite/page.html), but nothing that
contains queries (eg. /mysite/search.html?q=foobar). That gives a good
rule for programming your application.

Regards,
Alex

-- 
Alexander Klimetschek
alexander.klimetschek@day.com