Re: [Handle-info] redundancy questions

[Sean Reilly <sreilly@cnri.reston.va.us>]

On Nov 30, 2007, at 3:34 PM, consultmdb@nypl.org wrote:
> > The URL for the handle server process shouldn't be significant and I
> > would advise against directly addressing the handle server's HTTP
> > interface for any reason other than testing/debugging.  The handle
> > server is contacted by handle clients using its IP address (not  
> > domain/
> > host name) primarily using the native handle protocol over port 2641
> > (UDP and TCP with a fallback to HTTP).
>
> I'm thinking here primarily of non-handle-protocol-aware clients...  
> There's
> no way to embed a handle in a Web page without using a HTTP URL, is  
> there?
> And then doesn't using the domain name make more sense than an IP  
> address?
> Our ISP is far more likely to change our IP address allocation then  
> we are
> likely to lose our domain.

Since most clients are not aware of the handle protocol, http-based  
resolution is probably better off going through http://hdl.handle.net/ 
<your_handle> which will accept the HTTP request, resolve the handle,  
and send an HTTP redirect back to the client.  It is similar to the  
HTTP interface built into each handle server but has additional  
features, distributed load balancing, etc.  In addition, it is  
possible (hopefully likely) that future web clients will be able to  
recognize http://hdl.handle.net/* URLs and process them as native  
handle resolutions.

One benefit of the handle system is that it provides a framework for  
persistent identifiers.  Encouraging people to access them through a  
domain name negates that benefit.  Yes, hdl.handle.net is also a  
domain name, but one that has a strong assertion of being available  
long term (so that you can go on holiday and not worry about one of  
your servers crashing).

Thanks,
Sean



_______________________________________________
Handle-Info mailing list
Handle-Info@cnri.reston.va.us
http://www.handle.net/mailman/listinfo/handle-info