[Pipet Devel] node-to-node communication: you can have it 3 ways!

J.W. Bizzaro bizzaro at geoserve.net
Wed Mar 22 09:33:50 EST 2000

jarl van katwijk wrote:
> Chart 1 is not acceptable for security and management reasons,

(Chart 1: GMS brokers Overflow-to-Overflow subnet communication)

But *IF* Overflow can can be made secure and managable across the Internet,
then Chart 1 is an option, right?  I mean, it's just like Chart 2, except
Overflow is the 'foreign app'.

> chart 2 is fine, applications communication just the way
> they should have done without any system wrapping them,

(Chart 2: GMS/Overflow broker foreign app communication)

Right.  No arguments here.

> chart 3 is best when the internet had unlimited capacity :)

(Chart 3: No brokering; all communication goes through GMS)

Remember though that all those 2-way communication lines below GMS are kept
within the local host.  There's still only one communication line through the
Internet: between GMS instances.

> I like to see 2 and a 3b possible, where 3b is Jeffs 3,
> only the 'gms layer' will try to clone remote nodes to
> local addres space

Well, if the nodes are small enough, and the cloning process takes less time
than Internet communication, then why not?

> Any reason why 1 should not be thrown away?

If we agree that Overflow will not communicate with anything across the
Internet, then 1 can be thrown away.  I just wanted to leave that option open
to Jean-Marc if he wants to develop an Internet API later on.

                      |           J.W. Bizzaro           |
                      |                                  |
                      | http://bioinformatics.org/~jeff/ |
                      |                                  |
                      |        BIOINFORMATICS.ORG        |
                      |           The Open Lab           |
                      |                                  |
                      |    http://bioinformatics.org/    |

More information about the Pipet-Devel mailing list