You are viewing a plain text version of this content. The canonical link for it is here.

Posted to users@jackrabbit.apache.org by Marc Herbert <Ma...@gmail.com> on 2010/07/14 11:42:22 UTC

SPI versus RMI versus... SPI+RMI

Hi all,

  I am currently investigating remoting options. A number of
pages/posts state that there are two remoting options: "SPI" versus
"RMI". By digging deeper I think I found talk about a third option:
SPI "on top of" RMI (instead of WebDav). Is this correct? I mean, is
this actually a third option or am I just confused?

If there is indeed a third SPI+RMI option then it does not seem much
advertised. Is it deprecated or dead?


Performance aside, forwarding SPI method calls using RMI feels
more straight forward to me than going through an additional
WebDav/HTTP layer. Since RMI ignores the Fallacies of Distributed
Computing, remoting using RMI requires very little development
doesn't it?

How naive is the above?


Cheers,

Marc

Re: SPI versus RMI versus... SPI+RMI

Posted by Angela Schreiber <an...@day.com>.

hi marc

> Could you elaborate a tiny bit? The email archives would be grateful
> to you :-)

i guess marcel as the original author of the RMI-SPI code base
would be the best person to elaborate about the details. i only
adjusted minor things here and there without having a closer
look at the details... i just wanted to stress out the fact that
this piece of code never made it out of the sandbox and hasn't been 
adjusted to match the SPI 2.0. i assume it's not heavily used up
to now...

angela

Re: SPI versus RMI versus... SPI+RMI

Posted by Marc Herbert <Ma...@gmail.com>.

Le 14/07/2010 11:43, Angela Schreiber a écrit :

>>  i agree, remoting the SPI using RMI should be straightforward,
>> e.g.
>>   jcr2spi <-> SPI RMI client stub <-(network)-> SPI RMI server <-> spi2jcr
>>
>> AFAIK, nobody's implemented yet.
> 
> marcel did, didn't he?
> -> it's in the jackrabbit sandbox but not updated to the JR 2.0.

Stefan, Angela: thanks for answering.

Coming back to performance, would this "third" remoting solution be
usable over WAN latency? (unlike plain RMI).  I mean would calls be
batched somehow? This specific quote got me wondering:

  "While Jackrabbit SPI to JCR and SPI to WebDAV really are repository
   implementations, the SPI-RMI is rather a 'repeater' for SPI calls
   using RMI"
 http://jackrabbit.apache.org/jackrabbit-spi.html


>     and it's not optimized as far as i know.

Could you elaborate a tiny bit? The email archives would be grateful
to you :-)


Cheers,

Marc

Re: SPI versus RMI versus... SPI+RMI

Posted by Angela Schreiber <an...@day.com>.

hi

>> How naive is the above?
> 
> not at all ;) i agree, remoting the SPI using RMI should be straightforward,
> e.g.
>   jcr2spi <-> SPI RMI client stub <-(network)-> SPI RMI server <-> spi2jcr
> 
> AFAIK, nobody's implemented yet.

marcel did, didn't he?
-> it's in the jackrabbit sandbox but not updated to the JR 2.0.
    and it's not optimized as far as i know.

regards
angela

Re: Jackrabbit remoting possibilities

Posted by Marc Herbert <Ma...@gmail.com>.

Hmmm... mailing list stripping tiny attachments without a warning? Not nice.

I have uploaded the diagram here:
  http://wiki.apache.org/jackrabbit/RemoteAccess?action=AttachFile


Le 16/07/2010 16:19, Marc Herbert a écrit :
> Hi,
> 
>  Please find attached a diagram showing all remoting solutions as I
> understand them (it is partially based on
> http://jackrabbit.apache.org/jackrabbit-spi.html )
> 
> Could knowledgeable people please point out any
> errors and omissions?
> 
> Once corrected, would it make sense to attach this diagram to this wiki
> page? http://wiki.apache.org/jackrabbit/RemoteAccess
> 
> Thanks,
> 
> 
> Marc

Re: Jackrabbit remoting possibilities (was: SPI versus RMI versus... SPI+RMI)

Posted by ChadDavis <ch...@gmail.com>.

I don't see an attachment.

On Fri, Jul 16, 2010 at 9:19 AM, Marc Herbert <Ma...@gmail.com> wrote:
> Hi,
>
>  Please find attached a diagram showing all remoting solutions as I
> understand them (it is partially based on
> http://jackrabbit.apache.org/jackrabbit-spi.html )
>
> Could knowledgeable people please point out any
> errors and omissions?
>
> Once corrected, would it make sense to attach this diagram to this wiki
> page? http://wiki.apache.org/jackrabbit/RemoteAccess
>
> Thanks,
>
>
> Marc
>

Jackrabbit remoting possibilities (was: SPI versus RMI versus... SPI+RMI)

Posted by Marc Herbert <Ma...@gmail.com>.

Hi,

 Please find attached a diagram showing all remoting solutions as I
understand them (it is partially based on
http://jackrabbit.apache.org/jackrabbit-spi.html )

Could knowledgeable people please point out any
errors and omissions?

Once corrected, would it make sense to attach this diagram to this wiki
page? http://wiki.apache.org/jackrabbit/RemoteAccess

Thanks,


Marc

Re: SPI versus RMI versus... SPI+RMI

Posted by Stefan Guggisberg <st...@day.com>.

hi marc

On Wed, Jul 14, 2010 at 11:42 AM, Marc Herbert <Ma...@gmail.com> wrote:
> Hi all,
>
>  I am currently investigating remoting options. A number of
> pages/posts state that there are two remoting options: "SPI" versus
> "RMI". By digging deeper I think I found talk about a third option:
> SPI "on top of" RMI (instead of WebDav). Is this correct? I mean, is
> this actually a third option or am I just confused?
>
> If there is indeed a third SPI+RMI option then it does not seem much
> advertised. Is it deprecated or dead?
>
>
> Performance aside, forwarding SPI method calls using RMI feels
> more straight forward to me than going through an additional
> WebDav/HTTP layer. Since RMI ignores the Fallacies of Distributed
> Computing, remoting using RMI requires very little development
> doesn't it?
>
> How naive is the above?

not at all ;) i agree, remoting the SPI using RMI should be straightforward,
e.g.
  jcr2spi <-> SPI RMI client stub <-(network)-> SPI RMI server <-> spi2jcr

AFAIK, nobody's implemented yet.

cheers
stefan

>
>
> Cheers,
>
> Marc
>
>
>
>