[Biocorba-l] SOAP<->CORBA bridge

[Dickson, Mike]

See below:

> -----Original Message-----
> From: Philip Lijnzaad [mailto:lijnzaad@ebi.ac.uk]
> Sent: Tuesday, May 15, 2001 8:05 AM
> To: Matthew Pocock
> Cc: Ewan Birney; biocorba-l@biocorba.org
> Subject: Re: [Biocorba-l] SOAP<->CORBA bridge
> 
> 
> On Tue, 15 May 2001 12:44:09 +0100, 
> "Mat" == Matthew Pocock <mrp@sanger.ac.uk> wrote:
> 
> Mat> Very cool stuff - SOAP solves the granularity of 
> transport issues we 
> Mat> have (esp with *large* sequences) - and this bridge 
> looks very funkey. 
> 
> You solve the granularity/round trip problem, but you get 
> another problem in
> return; reports have it that SOAP traffic is likely to be 10 
> - 1000 larger
> than IIOP (especially for complex models). So some kind of 
> compression will
> be needed. But maybe it's easy to add a Content-Encoding: 
> gunzip or so, that
> would already make a big difference (although a factor of 1000 remains
> steep). 

I believe there are already a number of proposals for compression.  Actually
the content doesn't change so its really more an issue of running the soap
messages over a compressable transport that both ends agree on.

Regarding the size of the soap traffic it obviously depends on what your
sending.  If the traffic lends itself to being binary then clearly the
marshalling will need to stringinfy things and that can be expensize.  OTOH,
if your sending character strings then the I'd expect the cost to be
minimal.  I'd imagine in this domain we'll see alot of character data so
maybe its not so bad...

There are 2 big differentiatiors between IIOP and SOAP to me.  SOAP is
internet friendly (i.e. it tends to go through firewalls easier due to the
oft used http transport).  On the downside, its really just an RPC protocol
so there are no mechanisms (unless you invent one) to send objects by
reference.  Many of the things we'd likely want to implement probably lend
themselves to a services model (DAS, etc) so that may be a mute point.

> 
> Mat> Also, at that point, we don't need to make decisions 
> about which to use, 
> Mat> as they can shair an object-model.
> 
> Mat> Perhaps we should realy have a BioUML that is then 
> projected as CORBA, 
> Mat> SOAP etc. and transduced into the apropreate 
> language-specific bindings 
> Mat> for perl, java, python. 
> 
> yes, very much; this is also the direction OMG is heading (in 
> case anyone is
> intereseted :-), with their new Model Driven Architecture (with UML
> gradually replacing IDL)

I think the MDA is a *really big deal*.  It seems like were blessed/cursed
with new middleware and language technology about every 3-4 years.  Focusing
on a implementation nuetral model lets us translate to the implementation
technology of the day (reaping the presumed benefits) without the need to
build complex bridges that translate concepts semantically as well as
between the technologies feels like a big win to me.

Mike

> 
> -- 
> If you have a procedure with 10 parameters, you probably 
> missed some. (Kraulis)
> --------------------------------------------------------------
> ---------------
> Philip Lijnzaad, lijnzaad@ebi.ac.uk \ European Bioinformatics 
> Institute,rm A2-08
> +44 (0)1223 49 4639                 / Wellcome Trust Genome 
> Campus, Hinxton
> +44 (0)1223 49 4468 (fax)           \ Cambridgeshire CB10 
> 1SD,  GREAT BRITAIN
> PGP fingerprint: E1 03 BF 80 94 61 B6 FC  50 3D 1F 64 40 75 FB 53
> _______________________________________________
> Biocorba-l mailing list
> Biocorba-l@biocorba.org
> http://www.biocorba.org/mailman/listinfo/biocorba-l
>