Status D-AG0019: reliable, stable, and predictably evolvable Web Services

Goal statement [1]:
"reliable, stable, and predictably evolvable web services"

Strawman defns. of reliability and stability were presented in [1],
and some good discussions ensued. The "stability" defn. seems to be
fine, though "reliability" defn is being worked on. A defn.
for "predictably evolving ws" is still (predictably?) evolving.

Should we require "X amount" of reliability for all Web Services?
The consensus seems to be "no" [2], though definition and recommended
practices are within scope. In [6] there are arguments on why some
reliability requirements should be in scope.

[3] describes how a stable WS can evolve predictably. 
[4] suggests that, again, requiring stability of web services is not OK.
There is an idea that web services should use meta description
of the services to describe their stability and predictable evolution [4].
In [5], there is a suggestion that WSDL should
describe when the interface will be invalid ("time-to-live guarantee").
[5] also proposes that a service dispatcher could remember the service
versions instead of the services.

Though stable as of today, there are no reliability guarantees apart from
"best effort" for this summary. Further, YMMV on the predictable evolution
of this summary:-))

Regards,
-Suresh 




[1] http://lists.w3.org/Archives/Public/www-ws-arch/2002Mar/0309.html
[2] http://lists.w3.org/Archives/Public/www-ws-arch/2002Mar/0357.html
[3] http://lists.w3.org/Archives/Public/www-ws-arch/2002Mar/0351.html
[4] http://lists.w3.org/Archives/Public/www-ws-arch/2002Mar/0361.html
[5] http://lists.w3.org/Archives/Public/www-ws-arch/2002Mar/0366.html
[6] http://lists.w3.org/Archives/Public/www-ws-arch/2002Mar/0363.html

Received on Monday, 25 March 2002 19:32:44 UTC