Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
[geclipse-dev] Dynamic extension of a VO

Hi gEclipse folks

This mail should be the beginning of a discussion regarding the dynamic extendibility of the gEclipse framework. What do i mean by that? While providing an AWS EC2 implementation for gEclipse it became very obvious that the following structure is to be desired due to the nature of the Amazon webservice structure:

One AWS Virtual Organization acts as an umbrella facility to house many different services. Services such as EC2, S3, DevPay etc... All these AWS services are access through webservices and have nothing in common thereby. They act as individuals and although they are able to access each other (also internally) they do not share a common denominator. This implies a dynamic structure on the side of gEclipse. A small core providing the AWS VO and a set of protocol implementations bundled under the AWS VO.

Here is the problem. How is something like this achieved in gEclipse? As far as i know there is no congruent concept available. What i have seen are proprietary extensions to the VOMS vo which lets one add an url to a distant service. Gria uses the ability to add multiple data stagers and such. Theses approaches do not add new functionality but just a set of remote "data sources/webservices" which the implementation in geclipse already knows how to handle. Therefore this scenario is not feasible for the AWS integration, which wants to add new functionality.

Is there any other way to implement such a scenario as outlined above in gEclipse? Anything i am not aware of to add functionality dynamically? The most common and satisfying approach for the AWS integration would be the Eclipse concept of extension points:

The AWS plugin offers a set of extensionpoints to which an implementation of an AWS service (EC2, S3...) could be plugged. This would be the most generic way of integrating the various services under the AWS umbrella VO. Although this solution would be specific to AWS it is the most direct as far as i see.

I hope to get any feedback for this architectural problem. Especially the people in the core are advised to answer... mathias, ariel, harald? :)

thanks for any input.
regards
Moritz




Back to the top