> -----Original Message-----
> From: SRU (Search and Retrieve Via URL) Implementors
> [mailto:[log in to unmask]] On Behalf Of Mike Taylor
> Sent: Wednesday, April 14, 2010 3:18 PM
> To: [log in to unmask]
> Subject: Re: Add "reliability" index to CQL's "zeerex" context set
>
> On 14 April 2010 23:06, Peter Noerr <[log in to unmask]> wrote:
> > Leaving aside the point (for the moment) of *what* is reported; it is
> clear that Source uptime/reliability must be measured from an outside
> point (or points). However the Zeerex information is returned by the
> Source itself in response to a request.
>
> Not (necessarily) so. From the earliest days of the ZeeRex work it
> was envisaged that databases of ZeeRex records would exist, describing
> services other than themselves. See
> http://explain.z3950.org/overview/index.html#4
> for some background, and
>
> http://irspy.indexdata.com/doc.html?module=ZOOM/IRSpy/WebService.pod
> for an example.
>
>
My concern is not the existence of a Zeerex server to supply this service (which would be a great idea), but the management and maintenance of the service itself, so that it does not slip into the <95% availability, going down to 80%, now it's off because the power was cut....etc. Who, or which organization, is going to commit to running such a service. You have a very good one running in IRSpy. I didn't count the entries, but there must be a couple of thousand, so it has the coverage. But is IndexData going to provide a service that everyone can rely on? It could, but would the company commit to it? We do run such a service for our installations, but the coverage of "z" Sources is probably less than a thousand (the large majority are "html/???" Sources), and it is a service for our commercial customers, and I know what it takes to keep it up and running.
My thought about the "mirror" capability was so that all SRU/Z39.50 servers could run this and allow others to gather their own statistics if they wanted to, bouncing the traffic off friendly servers (multiple to cover for downtime). That way there is no central site, no management, no single point of failure, no registering, no bureaucracy. All round simpler to maintain.
|