Dear Steve Loughran, Many thanks for the software link and related information. The main focus of the GLARE is to provide dynamic mapping between abstract (semantic or functional) description to concrete (deployment) descriptions of a software component. For on-demand deployment, initially we use autoconf steps like 'configure|make|make install' or 'ant build|ant deploy', and did not use any special language for it, but this was not sufficient for legacy applications which sometimes require special env to be set and/or user interaction during the installation. In order to overcome this problem we introduced this simple language in which provider can provide send/expect like steps, which are executed automatically before on-demand provisioning. We would like to adapt ggf standards as much as possible. Currently we are exploring CDL and related APIs and will come back with some feedback. Thanks again. best mumtaz Loughran, Steve wrote:
Everything is under SCM; see https://sourceforge.net/projects/deployment for stuff to check out.
-you will need a prerelease of Ant1.7 to build the JARs, grab one from http://people.apache.org/~stevel/dist/ if you don't want to build one yourself
-We also build nightly on apache gump; if your project also has a public CVS or SVN repository, you can get added to the Gump and so catch any changes in CDDLM or any other Java dependency.
Regarding the GLARE language versus CDL, there is some appeal to Glare. Indeed, the Ant dev team is always bemused when people use it as the reference language for workflows; yours is probably the third such version, following on from GridAnt, and Sun's XML Pipeline language (http://www.w3.org/TR/2002/NOTE-xml-pipeline-20020228/). One thing that Ant does model well is the notion of files, paths and directories, though even there its model of resources is weak because the lack of unity across the many groups of things that you can use (hence ant1.7's refactoring to have an abstract ResourceSet as the parent of all groupings).
CDL tries to be more abstract, but personally, file systems are so important to deployment/configuration that its important to have a good way to model them. We end up having components which do nothing but map a platform independent filename into a platform-specific one, setting a lazy attribute to the final, absolute value at deployment time.
Tschuss, -Steve
-----Original Message----- From: Milojicic, Dejan S (HP Labs) Sent: 21 February 2006 14:08 To: Mumtaz Siddiqui; scrumb@ggf.org Cc: Loughran, Steve; Alex Villazon Subject: RE: CDL schema/API
Hi Mumtaz,
Sorry for the confusion, I was suggesting that you talk to Steve Loughran (cc-ed), not Steve Crumb.
Thanks,
Dejan.
-----Original Message----- From: Mumtaz Siddiqui [mailto:mumtaz@dps.uibk.ac.at] Sent: Tuesday, February 21, 2006 4:55 AM To: scrumb@ggf.org Cc: dejan@hpl.hp.com; Alex Villazon Subject: CDL schema/API
Dear Steve Crumb,
I presented my work about GLARE in first session of CDDLM at GGF16. GLARE provides XML based software component description and
automatic
deployments framework.
Currently we are using simple XML documents for software component descriptions, but as Dr. Dejan Milojicic suggested, we
would like to
use CDL for descriptions and deployment specifications
sometime later.
I could not find CDL schema and APIs (including Java stubs)
on CDDLM
homepage. May I request for a copy of schema as well as APIs?
Thanks with best regards mumtaz