[MOBY] [MOBY-l] Interpreting results from Service->execute()

[Ken Steube]

I don't think it's appropriate to delete people's services because they
didn't update the Time To Live. Let me suggest an alternative:

What I was trying to say in my last post was that I think it's more useful
for Central to store a "Time of Last Successful Test" for each service.
Service authors then could write a test for each service that upon success
updates this timestamp at Central. This test is what should go in cron for
periodic execution.

Then we could adjust findService() to have an extra option like
"testTime=>7" to request services that have been successfully tested in
the last 7 days.

I will also suggest another timestamp for each service: one that says the
service is known to be down as of a certain date, and the author has not
been able to fix it yet. With this you can notify the world your service
is down and you haven't fixed it yet.

With this approach nobody's service gets deleted and you have a really
useful way to see if a service is being maintained.

Ken

On Wed, 17 Mar 2004, Martin Senger wrote:

> > I wonder if this would be effective.  My approach to the Time To Live
> > approach would be to write a cron job that notifies MOBY-Central that my
> > services are still OK.  So my services might break and Time To Live
> > wouldn't be useful.
> >
>    You can always cheat - and break things, of course. But why to write a
> cron job that notifies if there is nothing to notify? :-) Better to create
> a cron job that can re-register my service after checking that my service
> is okay. The beauty of this is that the cron job is create by the same
> person that wrote the service so he/she knows the best how to check the
> service.
>
> > I like the idea of having a standard test script that gets run
> > periodically by someone's computer. It would have a carefully chosen input
> > to produce a predictable output which can be verified to be correct by the
> > test script.
> >
>    There is no such think as a standard test...
>
>    Just my 2c, Martin
>
>

-- 
--
--
Ken Steube
San Diego Supercomputer Center
University of California, San Diego, MC 0505
9500 Gilman Drive
San Diego, California, 92093-0505  USA
FAX (858) 822-3610