[mythtv-users] Mooting architecture for a DataDirect replacement

David Brodbeck gull at gull.us
Tue Jun 26 17:15:53 UTC 2007


On Jun 26, 2007, at 5:50 AM, ICMan wrote:

>
> The solution with the best model is DNS.  I am not suggesting we  
> use DNS servers and add an extra set of record types (though that  
> has some merit), but using the DNS model (and maybe a tight server  
> as the base of the code, like tinyDNS or somesuch) allows a  
> distributed heirarchy of servers, a common transport format,  JIT  
> information updates (depending on the timeouts on your local  
> server), the ability for redundant data storage - for huge numbers  
> of transactions consisting of relatively small amounts of data  
> (most hits will be differential data), DNS rules the Internet.   
> Literally.  With the properties of such a mechanism, why is it not  
> being seriously considered on this list?

It's just a bad fit, IMHO.

DNS is designed for quickly looking up small amounts of ephemeral  
data while actively online.  It excels at this.

But we need to download larger amounts of data in bursts and store  
it.  It's more of an issue of mirroring than database look-up.

Remember, not all Myth boxes have continuous Internet connections,  
and we shouldn't use a system like DNS that would require one.




More information about the mythtv-users mailing list