On 14.06.2006, at 21:06, Matthew Graham wrote:
> 11. Mandatory protocols: We discussed this back in October and
> could not actually come up with a list so I think that you have to
> stick with a list of suggested protocols and hope that most
> implementations cover some. I'm working on an implementation that
> uses http but also has support for jparss.
We cannot use "hope" as the basis for a specification - we have try to ensure that this specification guarantees interoperability as far as possible, or the specification is worthless. As I have already pointed out, with the current state of the specification, I can produce a compliant VOStore implementation and Dave can produce another, but because we have different ideas about the exact forms of transport mechanisms, they would not be able to transfer data between each other!
At a minimum we need to mandate one transport mechanism (and describe in detail what we mean by it) to ensure that compliant services can at least perform this rather basic use case. Given that we are defining a web service, http would not be a burdensome requirement for people. I also think that we should define a list of names for other common transport mechanisms - e.g. ftp, gridftp so that if two implementations decide to provide a gridftp transport they can interoperate because they both call it gridftp rather than one calling it something like "globusftp".
Paul. Received on 2006-06-16Z10:15:47