You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@directory.apache.org by Emmanuel Lecharny <el...@gmail.com> on 2005/11/21 15:02:16 UTC
Re: [mina] Object serialization integration to ByteBuffer (Re:
Decoding/Encoding UTF representation)
This is something that we can work on.
The main problem with this approach is that we have to describe the way
some objects are to be serialized. We can't rely on standard
serialization of HashMaps, Strings and aother java structures.
The second point would be to choose the best encoding solution. BER is
pretty verbose compared to PER, but as we won't benefit from many of the
PER advantages, this is not a real problem.
Last, I'm not sure that using ASN.1 against standard serialization will
really lead to improvment in performance. This has to be checked.
--Emmanuel
On Mon, 2005-11-21 at 20:25 +0900, Trustin Lee wrote:
> Hi Alex,
>
> 2005/11/19, Alex Karasulu <ao...@bellsouth.net>:
> Wish we had a nice little toolkit for using ASN1 to serialize
> objects
> and deserializing them. This would be a great place to apply
> it.
> Marshalling and demarshalling would be fast on the wire.
>
> I thought about this. The biggest problem is that the default
> ObjectInput(Output) stream is too slow. It is because we have to
> create the stream every time we call putObject or getObject. It is a
> huge performance hit IIRC from my previous projects. So I extended
> ObjectInput and Output stream for special purpose.
>
> The idea is very good in general, but we have to find out the best way
> to minimise the overhead. I'll post a JIRA issue for this.
>
> Trustin
>
> --
> what we call human nature is actually human habit
> --
> http://gleamynode.net/
Re: [mina] Object serialization integration to ByteBuffer (Re: Decoding/Encoding UTF representation)
Posted by Trustin Lee <tr...@gmail.com>.
The another problem with object serialization is that it is coupled with
internal JDK implementation. I think we can start from well-know object
serialization frameworks such as XStream.
Trustin
2005/11/21, Emmanuel Lecharny <el...@gmail.com>:
>
> This is something that we can work on.
>
> The main problem with this approach is that we have to describe the way
> some objects are to be serialized. We can't rely on standard
> serialization of HashMaps, Strings and aother java structures.
>
> The second point would be to choose the best encoding solution. BER is
> pretty verbose compared to PER, but as we won't benefit from many of the
> PER advantages, this is not a real problem.
>
> Last, I'm not sure that using ASN.1 against standard serialization will
> really lead to improvment in performance. This has to be checked.
>
> --Emmanuel
>
> On Mon, 2005-11-21 at 20:25 +0900, Trustin Lee wrote:
> > Hi Alex,
> >
> > 2005/11/19, Alex Karasulu <ao...@bellsouth.net>:
> > Wish we had a nice little toolkit for using ASN1 to serialize
> > objects
> > and deserializing them. This would be a great place to apply
> > it.
> > Marshalling and demarshalling would be fast on the wire.
> >
> > I thought about this. The biggest problem is that the default
> > ObjectInput(Output) stream is too slow. It is because we have to
> > create the stream every time we call putObject or getObject. It is a
> > huge performance hit IIRC from my previous projects. So I extended
> > ObjectInput and Output stream for special purpose.
> >
> > The idea is very good in general, but we have to find out the best way
> > to minimise the overhead. I'll post a JIRA issue for this.
> >
> > Trustin
> >
> > --
> > what we call human nature is actually human habit
> > --
> > http://gleamynode.net/
>
>
--
what we call human nature is actually human habit
--
http://gleamynode.net/