[Pipet Devel] Ideas for network distributed objects

Justin Bradford justin at ukans.edu
Fri Mar 17 00:45:57 EST 2000


http://www.levien.com/athshe/prelim-design.html

The above was written by Raph Levien, author of 
many things, including libart, Gill, and gdome.

Anyway, way back when I first got here, there was
lots of talk about how we'd access objects over
the network. I was proposing a DOM-like interface
over the network.

gdome, part of Gnome, was a DOM implementation
accessible through CORBA. Unfortunately, that's
really slow and there are lots of problems with
it.

So back when I was looking at these things, I
talked to Raph about gdome and we discussed the
possibility of a light-weight wire protocol to
do DOM like things over the network efficiently.

The link above describes Raph's thoughts/design
ideas on such a system.

Now, here is why I think this is extremely useful
for Loci (or whatever we call it now ;)

Bioinformatic/computational data can be very 
large, and it would be bad to have to transfer 
big blocks of data when you just need a bit of 
it.

Instead, imagine a simple API to a tree-like
structure. You could quickly browse through the
structure until you found the right nodes, 
lists, and/or blobs of data. Then you just get
the data you want. Even substrings of binary
blobs.

Also, it discusses the implementation of locking
and transaction for these distributed objects,
which could be also be very useful for things
like distributed computation and simulation.

Anyway, this Athshe describes my original vision
for distributed communication/object interaction
in Loci. I'm curious what people think, and if 
such a system would still be useful for Loci's
current direction.


Justin






More information about the Pipet-Devel mailing list