[MOBY-dev] Service deregistation *must* be fast: perhaps function call could precipitate prompt web agent visit?

Frank Gibbons fgibbons at hms.harvard.edu
Thu Sep 23 13:41:35 UTC 2004


Martin, Mark,

At 08:13 PM 9/22/2004, you wrote:
> >4) I use this list to strongly express how I agree with the last Rebecca
> >email about being still able to deregister a service by calling a
> >deregister method even if the service has been registered with the
> >signatureURL.
> >
> >
> >
>Well, I strongly disagree with you :-)   Gosh, THAT has never happened
>before!  ;-)
>
>You have to remember what the driving force was to create the RDF-based
>deregistration!! The problem is that any kid with the ability to read an
>API could have come in and desroyed our registry by deregistering all of
>the services in it with a 5-line Perl script!  Authentication was just a
>pain in the arse and totally impractical, so the only remaining solution
>was to remove that functionality of MOBY Central and move the problem to
>the service providers machine.  I am loathe to move away from that model
>because in the last 18 months nobody has come up with a better solution!!
>
>If a better solution exists, then we can certainly revisit the issue...

It seems to me that mostly people want rapid deregistration when they're 
actively developing a service. Once it's all set up, the need disappears. I 
have an idea, I don't know if it would be too much work, or too 
impractical, but here it is anyway....

What if different registries had the ability to set different 
deregistration policies. The 'real' Moby Central could allow dereg only by 
removing the RDF file, but the 'test' registry could allow this too, while 
also allowing dereg requests by API calls. I have absolutely NO idea how 
this might be implemented. But for me, I do my development work in the test 
registry, and only when I'm happy with things do I go ahead and put it up 
in the main registry.

Just a thought. I really must side with Rebecca - how can I debug a service 
if I have to wait days before the screwups I made in registering it are 
fixed by moby central's agent crawling over to my web page? Another idea 
might be that a function call causes the agent to come and look for the RDF 
within a short period of time (two minutes, say), and if it's gone the 
service is deregistered. I'm not advocating that the function call be 
blocking, of course, that would be impractical. It would always return 
quickly, and if there were anything returned, it could only indicate that 
the service had previously been registered, that the request had been 
received, and perhaps some estimate of how long before the agent could come 
by and check. There would be no indication of successful deregistration. 
That would require a separate call (to some other routine - findService?) 
after sufficient time had passed.

-Frank

PhD, Computational Biologist,
Harvard Medical School BCMP/SGM-322, 250 Longwood Ave, Boston MA 02115, USA.
Tel: 617-432-3555       Fax: 
617-432-3557       http://llama.med.harvard.edu/~fgibbons 




More information about the MOBY-dev mailing list