[Pipet Devel] Universal Description, Discovery, and Integration

[jarl van katwijk]

>
>
> Ooops.  We know how to describe and integrate :-)  Discovery, which will
> require directory services of some sort, is still kind of fuzzy to me.  I've
> been thinking along the line of how one searches Gnutella, as it is a similar
> P2P system.

Some time ago we agreed upon building a central registration server (on
bioinformatics.org at first). I think this will solve the discovery problem as
long as Piper remains unpopular. Once many people use Piper we'll be
screwed. Avoid too much advertising please ;)

If we were to build something beyond this limitation, we need a complex piece
of software called a hiarchial cache. Like squid can do. There are solutions to
the discovery prob., but they wont be easy.

I have something else that worries me more, let's keep up the atmosphere and
call it the pattern problem. It's about how to represent the executables. How
will
we layout the XML that describes the applications? I know we can stream
stdout and stdin data, we can build many 'templates' for a N number of
applications, but this will die a very chaotic death I'm afraid. I myself dont
know
what the best approach is to this XML patterning Piper will use. I'm very much
interested in your comment.

bye,
jarl