On Mon, 20 Nov 2006, Tamas Budavari wrote:
> Indeed, why not use S3, which BTW stands for Simple Storage Service!
>
> That is a nice summary, Matthew, and let me just add that most of these
> listed weaknesses of S3 may be considered its strengths, as well, and the
> rest (not REST!) is implementation details ;-)
Avoiding for the moment the issue of VOSpace vs S3; S3 of course does have a REST interface as well as SOAP; I don't know the statistics, but I'll bet 80-90% of the traffic is with the REST interface, as that is how these things generally run.
Seriously, SOAP is a nice technology, but like much new technology it is over-hyped. The main thing it gives you is a formally defined interface (via WSDL), and automatic language bindings - essentially IDL/CORBA for the Web. But this does come at a cost, and may not be cost effective unless you need to bind enough interfaces to make the cost worth the effort to invest in the technology needed to automate the language bindings.
> I recommend you read
> Matthew's message again, while chanting the following mantra:
>
> SIMPLE, WORKS, NOW :-)
>
> No, I am serious, read it again! People pay for Amazon S3... To me S3 is
> an interface definition that can be implemented to leverage the industry
> standard client tools (and all the documentation, user forums, etc out
> there) and there is nothing to stop us from extending its functionalities,
> having multiple instances of S3 servers, or even implementing other data
> access interfaces, such as VOSpace 1.x on top of the same data storage.
This argument probably holds true for many other things as well; WebDAV for example. Apple had something similar (although more proprietary) a while back. Others are sure to follow. For astronomy I don't think we really care how many folks use something like S3, mainly we just need something which works for us and is stable. I agree though, that it is always a good check to study these simple interfaces and ask why we can't use them, or at least what we can learn from them.