RE: Transactional Web Services (LONG)

[Nicolas LEHUEN <nicolas.lehuen@ubicco.com>]

Thank your very much for your answer. I'll need to read it once or twice
more to make sure that I got it correctly :), but I found it very
interesting.

I agree that the "contextual" web services we developed is a declination of
the client/server model. Indeed, it can be viewed as a transactional web
services model in which all transactional resources are found on the server
side. Yet, I still think that we should first solve the 2PC problem in the
"shop" model, because it should cover 80% (90% ?) of our current needs.

For now, I don't understand how you can expect two systems to interoperate
simply on the instance level, without relying on a set of schemas.

It made me think that in the "souk" model we would have to rely on what is
an old value in business, but a rather new one for IT protocol : trust.
Indeed, how do you guarantee that a transaction is effectively commited,
i.e. that your supplier will indeed ship the product because you paid him,
and conversely ? Do you think safe transactions can be passed without a
third-party guarantee, e.g. a bank certification service that proves your
supplier that you sent him the money, and a FedEx|UPS|ChronoPost|whatever
parcel tracking service that proves that the shipping has effectively be
done ? In the current transactional system, the client doesn't participate
to the transaction, has has to trust the server. Is there a way to change
this ?

Regards,
Nicolas

>-----Message d'origine-----
>De : W. E. Perry [mailto:wperry@fiduciary.com]
>Envoye : jeudi 23 aout 2001 21:45
>A : Nicolas LEHUEN; XML DEV
>Objet : Re: Transactional Web Services (LONG)
>
>

(see the archive for the original mail)