[Pipet Users] FW: Re: sbw

Jeff jeff at bioinformatics.org
Sat Oct 13 02:11:26 EDT 2001


>From Michael Hucka <mhucka at caltech.edu> (large pdf not forwarded):

-------------------------

  jeff> Hi Michael.  We just approved your project, sbw.
  jeff> Reading your description, I got the impression that
  jeff> the application has some of the same goals as Piper:

  jeff> http://bioinformatics.org/pipernet/

  jeff> I would like your opinion on where you see some
  jeff> differences, if any. Perhaps there can be some
  jeff> collaboration.

Hi Jeff,

Thanks for setting us up on bioinformatics.org.

We took a look at Piper some time ago and at the time, our
sense was that PiperNet was indeed similar but differed in
the following ways (and I admit these may be incorrect):

- The protocol seems to require XML in its data
  transmissions, whereas SBW supports a number of data
  types, both binary and string.  One *could* send XML using
  SBW, but it's not required.  We see this as an advantage
  because it permits fast binary data transmissions if
  desired.

- SBW is broker-based; this provides things like a registry
  and the ability to have callback notifications.  It's not
  clear if Piper uses a broker -- perhaps it does also?

- The SBW libraries for high-level object-oriented languages
  like Java provide proxy objects that greatly simplify the
  access of remote services.  In Java, for example, there is
  very little "SBWness" involved in getting applications to
  talk to each other.  We were trying to make SBW as
  transparent as we could to the application.  (Things are
  not so easy in C, unfortunately, where one ends up having
  to go through a lot more work.)

- Piper has a graphical dataflow-like layout system which
  SBW lacks.  Piper has a paradigm involving nodes and data
  flows between them.  These are cool ideas and advantageous
  in some context.  One point, however, is that although SBW
  elements could be viewed in a similar way, most of the
  tools for which SBW is intended are GUI tools and it's not
  so easy to program their interactions from a third
  entity.  Rather, the interactions emerge as users use the
  tools and switch from one to another.

However, there are definitely points of similarity between
Piper and SBW.  I think our group here needs to look at
PiperNet more closely.  I see that the PiperNet pages have
new material which we should look at.  

In case you're interested, I've attached a recent paper on
SBW.  This will appear in the upcoming ICSB2001 proceedings
(http://www.icsb2001.org/).  We have other documentation at

   http://www.cds.caltech.edu/erato

most of which will make its way to bioinformatics.org in
time.

A collaboration would be fruitful if we can find ways of
combining the best ideas of both systems.

Mike




More information about the Pipet-Users mailing list