dev@grizzly.java.net

Re: grizzly properties

From: Justin Lee <Justin.Lee_at_Sun.COM>
Date: Tue, 13 Jan 2009 10:19:08 -0500

Comments inline...

Jeanfrancois Arcand wrote:
> Salut,
>
> apology fo rthe delay. Swamped with fake memory leak ;-)
>
> Justin Lee wrote:
>> Any comments on the unused properties in the one pager
>> (http://is.gd/fxnq) before I start mapping these to xml elements?
>
> In section 4.12, if you can add number to table it might be easier to
> comment (ok very minor) :-)
>
> http-listener
> -------------
> external-port: not support
> family: not supported
> blocking-enabled: no longer supported with v3
>
> http-protocol
> -------------
> dns-lookup-version: not supported
> ssl-enabled: not supported
>
> http-service
> ------------
> http-file-cache: removed as well
>
>
> connection-pool
> ---------------
> receive-buffer-in-bytes: move this element under
> network-config>protocols>protocol>http. Might want to rename:
> request-body-buffer-in-bytes
> send-buffer-size-in-bytes: move this element under
> network-config>protocols>protocol>http
>
> file-caching
> -------------
> max-age-in-seconds:
> network-config>protocols>protocol>http>file-cache#max-age-in-seconds
> medium-file-space-in-bytes: move in
> network-config>protocols>protocol>http>file-cache#max-cache-size
>
> Properties to be dropped (Don't add them to the table as they don't
> need to be exposed as officially supported)
> ------------------------------------------------------
> com.sun.grizzly.enableCometSupport: in v3 it is called
> v3.grizzly.cometSupport
I don't see this property used anywhere in the grizzly source tree at
all. Not as listed and not even as simply "cometSupport." Are you sure
it's actually used?
> com.sun.grizzly.useDirectByteBuffer: remove
> com.sun.grizzly.pipelineClass: remove
> com.sun.grizzly.algorithmClassName: remove
> com.sun.grizzly.maxSelectors: remove
> com.sun.grizzly.factoryTimeout: ?remove, should be readTimeout
> com.sun.grizzly.asyncHandlerClass: remove
> com.sun.grizzly.asyncHandler.ports: remove
>
> readTimeOut/WriteTimeout: Keep it at the transport level IMO.
These are piped to static methods. If we have two transports that use
these elements, last write wins. Not very obvious, IMO.
>
> com.sun.grizzly.comet.notificationHandlerClassName : remove
> com.sun.grizzly.http.bufferResponse : remove
> com.sun.grizzly.OOBInline : remove *** we need to expose ALL socket
> properties the JDK is exposing. We probably need to add an element
> called <socket-property name="OOBInline" value="..."/>.
Is that a recommendation to add now? Or down the line? We already have
a <property> element. could be use that on the <transport>? I'm not a
huge fan of buckets like that as it doesn't make it clear what can and
can't be set but it certainly is a more flexible approach to tracking
Socket property options. Not that those change that much but...
> com.sun.grizzly.protocolFinders: remove
> com.sun.grizzly.protocolHandlers : remove
>
> Hope that help.
>
It does. I love deleting code. :)